The one feature of Redis I'd love to have supported in PostgreSQL is to be able to set a TTL on a record. On a recent project where we could only save personal data for so long, it was a must have feature so we had to use Redis for that purpose instead.
This would also give you audit logs, etc. As well as flexibility to adjust business logic without updating the TTL on all records.
BUT that comes at a cost. Doing that in Redis is just so easy, and the vast majority of the time no-one is going to notice a few concurrency errors, YOLO!
;-)
To be serious: it's a nice tool to have in the box and is amazing for inexperience/start developers because you can do a lot with very little time investment or knowledge.
Of course, at some point of scaling needs a dedicated cache store will make sense anyway.
(Just some justifications from a german engineer :) )
> Redis keys are expired in two ways: a passive way, and an active way.
> A key is passively expired simply when some client tries to access it, and the key is found to be timed out.
> Of course this is not enough as there are expired keys that will never be accessed again. These keys should be expired anyway, so periodically Redis tests a few keys at random among keys with an expire set. All the keys that are already expired are deleted from the keyspace.
> Specifically this is what Redis does 10 times per second:
1. Test 20 random keys from the set of keys with an associated expire. 2. Delete all the keys found expired. 3. If more than 25% of keys were expired, start again from step 1.
So really it's not much better than doing `SELECT value from keys where key=? and expires > now()` with manual deletion. Though I agree that it can be more convenient.
1. SELECT item_id WHERE expire = 0. If this is empty, no items are available.
2. UPDATE SET expire = some_future_time WHERE item_id = $selected_item_id AND expire = 0. Then check whether UPDATE affected any rows. If it did, item_id is yours. If not, loop. If the database has a sane optimizer it'll note at most one document needs locking as the primary id is given.
All this needs is a very weak property: document level atomic UPDATE which can return whether it changed anything. (How weak? MongoDB could do that in 2009.)
Source code at https://git.drupalcode.org/project/drupal/-/blob/9.2.x/core/... (We cooked this up for Drupal in 2009 but I am reasonably sure we didn't invent anything new.)
Of course, this is not the fastest job queue there is but it is quite often good enough.
I guess OP likes the simplicity that built-in expiration provides. In your example - all selects reading the value will need to have this expiration check. And also some scheduled process will have to be written to actually delete the values.
Don't use record lock as like it's an attribute of a row. Add a column on the table called "status" and use it.
(that being said, I try really hard not to judge; after all, i'm not without fault: it's 2021 and i'm using bash over cgi-bin to serve web pages for my own hobby projects :))) )
They use packages such as celery which happen use Redis under the hood but the user doesn‘t have to deal with these details other than pasting some Redis connection URL somewhere.
The main issue is making a job queue dispatcher and runner which most people aren't familiar with.
Any time someone says ‘this works for small workloads’ I hear, ‘we’ll eventually have to migrate this’.
That’s not a problem if you are talking about introducing a hugely complex piece of infrastructure, but Redis is hardly that.
Single user local applications? Fair.
Web applications? Very strange choice imo.
Reddis is great, but it is *not* a database, and it's thoroughly rubbish at high load concurrency without clustering, which is (still) a massive pain in the ass to setup manually.
Of course, you can just use a hosted version off a cloud provider... but, it's generally about 10x more expensive than just a plain old database.
/shrug
I mean, sure, it's (arguably...) step up from just using sqlite, but... really, it's easy, and that's good... but it isn't good enough as a general replacement for having a real database.
(To be fair, sqlite has got some pretty sophisticated functionality too, even some support for concurrency; it's probably a step up from redis in many circumstances).
By all accounts Postgres seems to be a pain to scale off a single machine, much more so than redis.
With Postgres you also need to worry about high churn, especially since you are creating/locking/deleting rows constantly. This can be alleviated through a variety of means, of which personally I would use per-day table partitioning and truncate older partitions on a cron, not to mention the sharp increase in database connections to the host now required.
Ignoring the literal elephant in the room of synced writes to the store. Redis can be used quite effectively in a blocking manner with RPOPLPUSH/LMOVE(6.2+) for a reliable queue, allowing an item to not be lost because atomically the pop and push from two different lists are done together.
[1] https://github.com/que-rb/que [2] https://gist.github.com/chanks/7585810
Additionally, the expire/ttl/get/set in Redis is incredibly easy to use (and abuse, hence the OP article). Some team's criteria is limiting the amount of moving parts - and that's great. Don't use Redis and use a relational database for everything such as what you mentioned. Use it as a queue, a cache, a message broker, etc..
Other teams may care less about an extra moving part if it means their code will look simpler and they leverage relational databases for their more common usecases.
Oh, and i'm not entirely sure about the part about redis active expiry (disabled by default, default is remove expired on lookup - lazy); you're talking about key eviction which applies to all deleted keys and AFAIR happens only when certain watermarks are hit. Since it happens in ram, it's also faaaast, unlike SQL DELETE, which will definitely involve disk...
Creating and deleting a lot of data on PG is a pain because of MVCC and vacuum. One useful trick is to partition data into tables and to truncate/drop entire tables, drop/truncate is instant and reclaims space immediately.
references:
0 - https://aphyr.com/posts/283-jepsen-redis
1 - https://martin.kleppmann.com/2016/02/08/how-to-do-distribute...
That being said, I've found Redis largely a pleasure to work with for these use cases and don't really see a real incentive to changing my current approach.
It's also got loads of complex and useful instructions.
1 - I replied in a blog post refusing most of the arguments, and then there was maybe another reply. You may want to read the full thing to form an idea about the safety of RedLock.
TTL on systems like Cassandra is pretty ugly and deleting data is hard as you scale no matter how you do it. I don’t think Postgres would be able to implement a TTL that is magically better than the rest.
You can also listen to Postgres WAL changes to build an Event like system with Elixir. Supabase is doing this with their Realtime app.
Isn't Redis an in-memory key/value store? Paying for RAM is not the same as paying for SSD.
If you're using postgres for caching only, as you do Redis, then you also do not need to provision block storage.
If you happen to already have Postgres running for other uses, you also do not need to provision block storage.
Finally, I would add that Redis clients such as Redisson are resource hogs that cause performance problems on apps, while pg clients are barely noticeable.
Also, most applications have peak and low periods of load that are predictable (e.g. users concentrated in a given set of time zones) which make for good times to run otherwise disrupting functions, etc.
You could also use a library like PG boss to handle the cleanup task.
Since you mentioned Cassandra and TTL, I'll mention ClickHouse, very nice TTL options, splitting into smaller partitions and using "ttl_only_drop_parts=1" has prove itself in the production with big data ingestion rates.
Last, but not the least, I almost always prefer Postgres for data storage needs, one can trust it to be safe and fast enough. Only some specific situations warrant other solutions, but it's a long way until that point (if ever), and better not optimize too early.
create table all_items(id integer, value text, expires timestamp);
create index all_item_expiry on all_items(expires);
create view items as (select id, value, expires from all_items where expires > now());
Then you can treat items as your base table and postgres neatly allows INSERT/UPDATE/DELETE from it. You'll need a job to clean up expires < now() items but it can be done at whatever arbitrary interval you like, could even be a trigger in PG if you were feeling spicy.
Disclaimer: I work on Debezium
That said, due to it's single-threaded nature, blocking on quorum writes is likely to bottleneck your application under any kind of significant load. It really shines at volatile data, and while it can work for valuable data, there are better tools for the job.
Basically you have status = 0,1,2 for queued, processing, done. You have a filtering CTE that returns the queued rows, sort by id and update status = status +1 of the top one.
You can even get fancy and disallow updates on the status column for other users to make it immutable.
I've done this in SQL server in a very similar way to as explained in the post
Personally I aim to develop with future migrations in mind. Certain abstractions are worth it. Put an interface around a cacher, db store etc. Day to day dev overheads is low, but your future self will thank you should you need to migratw away.
Sure this isn't always possible, but do it as much as possible.
Any opinions and experiences here with Kafka vs Redis as a queue?
As a capability, compare-and-swap has an infinite consensus number [1], meaning it's sufficient to implement wait-free consensus algorithms with an arbitrary number of participants. That makes it a perfect fit for managing a scalable pool of workers that need to coordinate on consuming from a queue!
[0] https://en.wikipedia.org/wiki/Compare-and-swap
[1] https://en.wikipedia.org/wiki/Consensus_(computer_science)#C...
In this case, the requirement is that user data must only be kept for a certain time and not longer.
If that time is a property of the record create and we're allowed to keep the data with an error of a day, I guess it's easy: We partition the table on a daily basis, and delete any partition that is older than 28 days old. Sometimes, a record will be closer to 29 days old when it's deleted, but we accepted that, and it's easy enough to write queries so that it's unavailable to the application if it's more than exactly 28*24*60*60 seconds old if that's our constraint.
If the requirement is to keep it based on the last of a certain kind of use, we'd need to move it from one partition to another if we need to keep it. For instance, if we can keep data for 28 days after the user last longs in, we can't just drop the partition - unless we've moved the user each day they log in.
If we have that kind of a constraint, where data lifespan is based on properties that change over the lifetime of the data, is partitioning + drop actually a useful approach? The drop will still be instant, I guess, but it's the movement of data over its lifetime that concerns me here.
It's so simple any little script can interact with it in seconds, instead of having to craft complex SQL or import your wrapper. You can call redis from the weirdest places, like from the inside of an nginx config file. You needn't even a compiled driver if that matters.
It's easier to get it to perform. It's possible to get great perfs with PostgreSQL, but you just got them for free with redis. Very hard to screw up. Read, write ? N+1 ? Batching ? Who cares, it will be fast no matter the concurrency or the load.
Sure you can expire with Postgres, but having done so in practice, it's way harder than it looks to get right. With redis, you don't have to care. Set a cache, set a lock with a timeout, store data your are not sure you need. It will all disapear.
Redis is not just key/value. You have sets, so ensuring unicity is a piece of cake, no constraints to define, then insertion to check against. And of course, you have sorted sets, which you are kept ordered at insertion by any number you pair the value with, such as a timestamp/score/index, and truncate by range for extra magic.
And then you have bonuses like hyperloglog which you need an extension for in Posgres. Super handy for stats.
Finally, you have streams, which for most apps will fill the gap for a timeserie database or a way to store your app logs. All that with a 2ms latency at 10k/s requests. None of my projects ever needed more than 1K/s though, even one with 1M users/month.
You have all of that with a dead simple install and basically no maintenance.
In fact, redis by itself consume almost no resource, it's not even worth it to not have it in your toolset. I just install it by default on all my projects: I never regretted it, there is always something it can do for you. It not now, just wait a bit, it's costing you nothing, and something will come up.
So no, let use Postgres for what it's great at, which is being a robust all purpose database. Redis is a fantastic complement to it, not a competitor, just use both.
Unless you are google size, there are little chances you will reach a stage where you need to migrate from any of them.
It's part of those tech that are just too good to be true, like SQLite or Python.
My only regret is that it doesn't exist on windows.
P.S: if you need caching and can't afford redis on a small python script, use Disk Cache, it's awesome: http://www.grantjenks.com/docs/diskcache/index.html
[0]: https://blog.2ndquadrant.com/what-is-select-skip-locked-for-...
This sounds like it might be vulnerable to dead consumers causing WAL to pile up without the right settings/periodic checks.
Which is probably why if you don't want loss if consumers go down a proper queue system (RabbitMQ, ActiveMQ, Amazon SQS, Kafka if you don't care about ordering between partitions) is the way I'd go.
Using a DB as a queue has been a thing for a very long time. Every billing system I've seen is a form of a queue: at a certain point in the month a process kicks off that scans the DB and bills customers, marking their record as "current".
The challenge is always going to be: what if the worker dies. What if the worker dies, the job is re-ran, and the customer is billed twice. Thank god it's been many years since I've had to touch cron batch jobs or queue workers. The thought of leaving the office knowing some batch job is going to run at 3am and the next morning might be total chaos... shudder.
Most of the time, you really don't need to scale postgres more than vertically (outside of the usual read replicas), and if you have tons of reads (that aren't hitting cache, I guess), then you can scale reads relatively easily. The problem is that the guarantees that postgres gives you around your data are research-level hard -- you either quorum or you 2pc.
Once you start looking into solutions that scale easily, if they don't ding you on performance, things get murky really quick and all of a sudden you hear a lot of "read-your-writes" or "eventual consistency" -- they're weakening the problem so it can be solved easily.
All that said -- Citus and PostgresXL do exist. They're not perfect by any means, but you also have solutions that scale at the table-level like TimescaleDB and others. You can literally use Postgres for something it was never designed for and still be in a manageable situation -- try that with other tools.
All that said, KeyDB[0] looks pretty awesome. Multithreaded, easy clustering, and flash-as-memory in a pinch, I'm way more excited to roll that out than I am Redis these days.
This is unfortunately the status quo in the slightly less ideal universe than “seemed like a good idea at the time”
Use SQS or RabbitMQ or something.
Edit: Also if you’re using something SQL for queues you’re going to have to build out your monitoring stack yourself based on your implementation rather than suck up a metrics endpoint with queue stats or pull from cloudwatch.
I am literally in the middle of digging a company out of this mistake (keeping Redis too long) right now. If your software/data is worth something, take a week or a month and figure out a reasonable schema, use an auto-generation tool, ORM, or hire a DB for a little bit to do something for you. Even MongoDB is better than redis if your're gonna do something like this.
* Every database a Postgres 1: Key/Value store
* Every database a Postgres 2: Document stores
* Every database a Postgres 3: Logs (Kafka-esque)
* Every database a Postgres 4: Timeseries
* Every database a Postgres 5: Full Text Search
* Every database a Postgres 6: Message Queues
Low key, you could make almost every single type of database a modern startup needs out of Postgres, and get the benefits (and drawbacks) of Postgres everywhere.
Should you do it? Probably not. Is it good enough for a theoretical ~70% of the startups out there who really don't shuffle around too much data or need to pretend to do any hyper scaling? Maybe.
If anyone from 2ndQuadrant/Citus/EDB see this, please do a series like this, make the solutions open source, and I bet we'd get some pretty decent performance out of Postgres compared to the purpose built solutions (remember, TimescaleDB did amazing compared to InfluxDB, a purpose built tool, not too long ago).
New features like custom table access methods and stuff also shift the capabilities of Postgres a ton. I'm fairly certain I could write a table access method that "just" allocated some memory and gave it to a redis subprocess (or even a compiled-in version) to use.
[EDIT] - It's not clear but the listing is in emacs org mode, those bullet points are expandable and I have tons of notes in each one of these (ex. time series has lots of activity in postgres -- TimescaleDB, native partitioning, Citus, etc). Unfortunately the first bullet point is 43 (!) bullet points down. If someone wants to fund my yak shaving reach out, otherwise someone signal boost this to 2Q/Citus/EDB so professionals can take a stab at it.
[EDIT2] - I forgot some, Postgres actually has:
- Graph support, w/ AgensGraph now known as AGE[0]
- OLAP workloads with Citus Columnar[1] (and zedstore[2]).
[1]: https://www.citusdata.com/blog/2021/03/05/citus-10-release-o...
The gist of the original article is asking whether you could reduce tech stack complexity and use a single set of technologies for more use cases, allowing to understand the tool you are using better.
Also, note that a traditional database may or may not be the right tool for the job - there are different storage/behavior needs for writing lots of data, processing lots of data and serving the results.
Unless we’re talking about a high load, there should be no problem doing this.
We have literally processed tens of billions of jobs without a single failure.
Old and simple technology that just works
It's also useful when you want to quickly build a "good enough" version of a feature like search so you can get it in front of your users fast and iterate on their feedback. Most of the time, they'd be quite happy with the results and you don't have to spend time on something like managing Elasticsearch.
I wrote a post on how you can use postgres to add search capabilities with support for queries like
jaguar speed -car
ipad OR iphone
"chocolate chip" recipe
http://www.sheshbabu.com/posts/minimal-viable-search-using-p...
That said, I am an unrepentant yak shaver, and there is a lot to be said in just writing those things when you need it but, Postgres would be perfect for rapid prototyping in this way.
Postgres, SQLite and many others are durable by default. Almost all so-called databases are like that. When you need a database, 90% of the time, you want durable. People make mistakes, developers are people, developers make mistakes, and one such mistake is assuming that Redis is like other databases in being durable by default when it's not. It's not conjecture, I've seen it done in production.
How so? What‘s wrong with SQLite?
More: https://use-the-index-luke.com/sql/where-clause/partial-and-...
For example Python has Celery and Ruby has Sidekiq. As far as I know there's no libraries in either language that has something as battle hardened with comparable features for background tasks using Postgres as a backend.
There's a big difference between getting something to work in a demo (achievable by skimming PG's docs and rolling your own job queue) vs using something that has tens of thousands of hours of dev time and tons of real world usage.
I'm all for using PG for things like full text search when I can because it drastically reduces operation complexity if you can avoid needing to run Elasticsearch, but Redis on the other hand is a swiss army knife of awesome. It's often used for caching or as a session back-end so you probably have it as part of your stack already. It's also really easy to run, uses almost no resources and is in the same tier as nginx in terms of how crazy efficient it is and how reliable it is. I don't see not using Redis for a job queue as that big of a win.
the point (ok, one point of many) of REDIS is that it is not the main DB so you can have a sense of security and decoupling in the architecture. Besides - there is no silver bullet for all things. While you can have your app do everything with PostgreSQL (and much more with Oracle, something people dislike it about), the fact itself does not mean is a good design decision or is a more stable decision.
Because when you have redis for sessions, kafka for event streams, postgre (or else) for data storage - you have components that can fail separately and thus the system degrades gracefully.
I come across a lot of SQL databases that uses 0 instead of false.
Is there any good reason for this?
- Postgres is probably already running (it's pretty good for OLTP workloads)
- Operational ease and robustness
- Cloud support everywhere for Postgres
- People know how to backup and restore postgres
- Sometimes Postgres will beat or wholly subsume your specialized system and be a good choice
- Postgres has ACID compliance and a very good production-ready grasp on the research level problems involved in transactions. I've never met an etcd/zookeeper cluster I didn't wish was simply a postgres table. Image being able to actually change your cache and your data at the same time and ensure that both changes happen or none of them happen (this is a bit vaporware-y, because locks and access pattern discrepancies and stuff but bear with me). You're much more unlikely to see Postgres fail a Jepsen test[0]
[0]: https://jepsen.io
Complexity comes at a huge cost. Only add it when the benefits out weigh the costs.
You could start out building a product that could scale to millions of uses overnight. But if you do that you've spent months building something with no users. You could have been in the market already, building revenue already. Your requirements will change as you find your market fit and you'll need to change things. The less you have built the easier it is to change. Why not leave the million user capabilities until you actually need it?
Sometimes it's an artifact of the orm mapping.
I drank from the “keep your queue out of the dB” koolaid once, but then you end up worrying about transaction boundaries and atomic writes to the queue and the dB. It totally depends on your workload, but in my case, I’ve found the halfway solution of the queue in the dB so you get to write to it within your transactions, with a dedicated queue for the workers to be a much better solution.
A good job queue will have all or most of these features:
Prioritize jobs (queues, weighs to certain jobs, etc.)
Scheduled jobs (running them once but X time in the future)
Periodic jobs (running them every 2nd Tuesday at 3:33am)
Static and dynamic configuration (CRUD'ing jobs at runtime, like adding new scheduled tasks)
Re-try jobs (customizable strategy, such as exponential back off)
Rate limit jobs
Expire jobs
Cancel jobs
Unique jobs
Batch executing
Handling graceful shutdown (integration with your app server)
Get metrics (status, health, progress, etc.)
Browse job history
Web UI (nice to have)
Rolling all of these things on your own would be a massive undertaking. Tools like Celery and Sidekiq have been actively developed for ~10 years now and have likely processed hundreds of billions of jobs through them to iron out the kinks.
Even if you managed to do all of that and created it as a Postgres extension (which I think is doable on paper), that's only half the story. Now you'd have to write language specific clients to interface with that so you can create jobs in your application using a nice API. This would be a very welcome project but I think we're talking a year+ of full time development time to release something useful that supports a few languages, assuming you're already an expert with writing pg extensions, have extension knowledge about job queues and know a few popular programming languages to release the initial clients.
Having the queue on its own means you have to handle exceptions like "job is done but queue didn't get updated" or get into distributed transaction which is very different challenge.
Also, operationally having one inherently stateful component (db) versus two (db, queue) makes things a lot simpler to manage, deploy, update, etc.
Not claiming this is a perfect solution, just that the tradeoff isn't obvious.
My clients seems rather happy with the project.
Whatever happened to use the proper data structures for the job? PostgreSQL and MySQL are, at the end of the day, b-trees with indices. Throw in relational properties and/or ACID too. Those aren't properties you need or want in a queue structure.
I know I don't have a solid argument against not doing it; it's just experience (and dare I say, common sense) telling me not to. Not quite like parent but I spent the first two years of my professional career in a team that had the brilliant idea to use DBs as queues. The big task I partook in for that stint is moving them off that v2 into a v3 which used---wait for it---Redis. Everyone's quality of life improved with every migration, proportional to the size of the v2 cluster we retired.
Is contention with the application your only objection? That’s pretty weak.
What’s the real compelling reason not to use a database? You haven’t said.
Im honestly unsure if you mean this as opposing "doing everything in Postgres" or as opposing "throw more services on the stack".
Because both are true for the statements. You have that complexity, regardless of where you implement it. Either you are building the rube-goldberg machine inside of postgres out of modules, config and SQL or outside of postgres with additional services.
The only way to really solve this is to avoid building that RG machine in the first place. In this case: don't have queues. In practice that probably means introducting complexity elsewhere, though.
Sorry but am I the only one who is very worried about the state of software? There are people who drank so much of the schemaless (which was not an actual issue for any dev worth her salt to begin with) that you have to dispense this kind of advice? I find that bordering on criminal if someone did that to you and carries the title programmer.
Again, maybe that is just me.
Edit: not an attack on the parent: good advice. Just didn't know it was that bad. And sad.
Not to mention numerous database administrators crying into their cups of coffee.
Enough?
Like I said 20 years of people fucking this up is my experience.
By putting it in the same transactional store with the same transactional boundaries you’re instantly shooting the whole point.
Not only that, most queues define boundaries between different logical systems where transactional boundaries do exist. At which point your database’s transaction scope is extremely limited in capability.
In the real world of messaging transactions span more than just the database engine.
It’s just bad architecture. Full stop.
Connection 1
LISTEN 'job-updates';
BEGIN;
INSERT INTO jobs ('a-uuid', …);
SELECT PG_NOTIFY('job-update', 'json blob containing uuid and state change info');
COMMIT;
BEGIN;
SELECT id, … FROM jobs WHERE id = 'a-uuid' FOR UPDATE SKIP LOCKED;
UPDATE 'jobs' SET state = 'step1_completed' WHERE is = 'a-uuid';
SELECT PG_NOTIFY('job-update', 'json blob containing uuid and state change info');
-- do the thing here: computation, calling external API, etc. If it fails then rollback.
COMMIT;
The only downsides I immediately think of are you will have every worker contending to lock that row, and you’ll need to write periodic jobs to cleanup/retry failures.
Databases actually work fine as a queue but emotionally you don’t like it. That’s fine it’s just not real strong objections.
What you have not said is “it physically does not work”, and that’s because it does work fine.
Honestly, I had more code managing two phase commits so I would only push to the queue when I was sure (ish) the dB commit was ok.
Say you need to trigger something to send an email, but let’s say that there’s an error condition after and the transaction is rolled back. How do you handle that failure? In my scenario the queue is rolled back too, no harm to foul.
The idea I was trying to get at was using redis to store data traditionally reserved for OLTP workloads.
> Pretty sure we all can think of some pretty high profile examples of NoSQL + structured data working very very well at scale.
Well that's the thing, you very rarely hear of companies who cursed their decision early on to use NoSQL when they realized that their data was structured but in 20 different ways over the lifetime of the product. Some datasets only need light structure (key/value, a loosely defined document, schema-included documents), and other things should probably have a schema and be stored in a database with a tight grip on that schema and data consistency/correctness. Please don't use redis in that latter case.
One huge advantage over, say, SQS, is that you also get easy visibility! You can query and see what’s been completed, what’s still not worked on, etc.!
Oh, and do you want a priority queue? Just add a priority field and sort before picking an item! Do you need to change the priority of an item while it’s already in the queue? Go right ahead. Do you want to enforce some constraints so that duplicates don’t get added to the queue? It’s a database — add the constraints.
If you’re already using a relational database, and now you need a queue, start with this approach. Odds are, it’ll work just fine, and it’ll give you the perks above.
Redis is, far and away, _the_ most robust piece of software I've ever had the pleasure of deploying to production.
This so much. People too often treat databases as magical black-boxes that should handle anything. Database is most often the bottleneck and choosing the proper storage engine with appropriate data structures can be 100x more efficient that just using the defaults. 1 server vs 100 can definitely make a noticeable difference in costs and system complexity.
While premature optimization is bad, choosing the right tool for the job is still somewhat important and will usually pay off in the long run.
As a result we use MySQL w/ memcached, although we are considering a swap to redis for the caching layer.
I think if I were going to argue against using DBs as queues it would be around: heavy parallel write use cases, latency concerns of both reads/writes and scaling to millions of events per second.
If you don’t have those concerns using a properly normalized and protected schema (which you are doing anyway right? Cause if not you are already shooting your toes off) for queues goes a very long way and removes a very big operational burden and tons of failure modes.
Frameworks like Laravel make this easy because they allow you choose different backends for your DB, queue, and cache, among other things, all or some of which can be an RDBMS like Postgres or MySQL/MariaDB.
When your app's need outscale what the RDBMS can rationally handle, then you can dedicate the resources to switch to a more specific tool as your growth has indicated you should.
I can't say what the percentage of apps that can handle things this way, but the anecdotes mentioned in this thread are clearly the cases where growth outstripped what a traditional RDBMS like Postgres are well-suited for, and so migrating was the right thing. The probably was likely identifying the need or delaying the transition until it was too painful. I don't think there's a magic bullet here, but I also don't think that starting with a dedicated queuing store tool is also always the right thing to do.
...but, I started writing about clustering and the network API, but, I can't really articulate why those are actually superior in any meaningful way to simply using sqlite, and given the irritation I've had in maintaining them in production in the past...
I guess you're probably right. If I had to pick, I'd probably use sqlite.
We were already using Redis for other things, it was the logical choice for this scenario.
Obviously depends on the scale and needs of a project, Postgres etc is fine for simple queues.
I often see people waste unnecessary time by writing their own complex solutions, resulting in increasing technical debt, when you already have perfectly suitable open source options available that do a better job..
Also I'd detect a worker has died by recording the start-time and using a timeout. Furthermore I'd requeue requests as distinct new entities. A requeued entity would have a self-referencing nullable FK to reference its parent request.
1) 99.9% of internet-facing/adjacent businesses are not Google and will never reach even 1% of Google's scale
2) Proto + BigTable is very different from just throwing stuff in redis/mongo. Proto schemas are compile-time enforced, which is great for some teams and might slow others down. Google enforces more discipline than your average engineering team -- this is overkill for most engineering teams.
You’d be trying to store a tiny bit of simple state and all the books/articles would have you standing up read only views & stored procedures for all your crud ops. The document stores came along with a fresh perspective and easy scaling.
Then their were the columnar stores and time-series stores that really did solve the newer scale problems in ways the existing sql stores didn’t.
I’m a sql guy through and through but it’s important to recognize the nosql movement was a reaction to real pain points. Also it made the sql databases better.
Postgres is awesome but logging and queuing in a table can cause gotchas you won’t have with redis.
I once was the one implementing this stupid idea. It's very easy to shoot yourself in the foot. It seems to be ok at first, but it is a stupid idea and will give nightmares to the operations team.
It works as long as the table is small. It gets especially problematic when there are sudden spikes in the workload, so that the table is growing rapidly. People don't test these scenarios enough.
Often it's important to do the jobs in the order they came in. This means you need to have the index set up this way or you'll end up with full table scans. You also need to remove jobs which are done, but this will also mess with the tables performance.
There are tricks with databases to help scaling this, but at the end of the day it's much simpler to just use a queue.
As always, there are trade offs, no silver bullets.
Do you mean "you'd probably already have indices by user then, so you won't be able to take advantage of quick drops"?
> There also other ways to store the data, for example not as a time series of user actions, but as another data structure. It just depends on your requirements and data model.
I suppose I want to be a bit more specific. Till now, if I wanted to deal with this issue, I would just have used a cron job and some indexes and taken on the load. But what is an example of a nice way of dealing with this? You get to make up plausible requirements and data model details, perhaps drawing on particular requirements you had when you faced a similar issue.
> The gist of the original article is asking whether you could reduce tech stack complexity and use a single set of technologies for more use cases, allowing to understand the tool you are using better.
The legitimacy of a question in a free-form conversation doesn't depend on the original inspiration, several iterations ago. But even if it did, the question is here is exactly about understanding the tools better and whether we can use one tool instead of two, so by your summary it's perfectly on topic.
> Also, note that a traditional database may or may not be the right tool for the job - there are different storage/behavior needs for writing lots of data, processing lots of data and serving the results.
The subquestion here is precisely if we can get away with just using Postgres for data with mandatory lifespan requirements that vary over the lifetime of the data.
Extra tools come at a cost - that is the presumption of this article and the thread it has spawned.
If we have to use Postgres, we need to assess the cost of doing this in Postgres before we can decide whether or not to pay the cost of other tools. Waving in the general direction of other tools isn't helpful; it probably takes as much work to enumerate candidates as it does to calculate the cost of doing it in Postgres.
Edit: can you explain by what you mean by solving it with topics? I thought topics were a broadcast mechanism?
In SQS, for example, you use a visibility timeout set high enough that you will have time to finish the job and delete the message before SQS hand it off to another reader.
You won’t always finish in time though, so ideally jobs are idempotent.
Of course, this relies on the jobs being something that can be retried.
That is definitely not ok. I'd be really pissed as a user if I wrote a huge comment and it suddenly disappeared.
Also in the cool-redis-stuff category:
https://github.com/twitter/pelikan
Doesn't have the feature set that KeyDB has but both of these pieces of software feel like they could the basis of a cloud redis product that would be really efficient and fast. I've got some plans to do just that.
How is this done in Redis automatically? At some point you are writing queries (albeit Redis ones) to pull metrics. Other queueing systems may expose an API but there is always some level of integration. With Postgres/other db I would write SQL which is their API and a powerful one at that.
I can couple events with triggers to auto-generate other events etc, have built in audit capability, roll up reporting etc, all with standard SQL. (all in the context of monitoring and reporting)
Going from a single database to a (database + queue) means two server processes to manage, maintain, observe, test etc.
I actually start with SQLite to reduce as much distributed state as possible, then move to something else once it’s proven it will not work.
1. It’s one tool to learn. No need to upskill on topics, partitions, consumer groups, routers, brokers, etc. 2. It’s one transaction boundary. You get a lot of “magic” for free when you know that a rollback will not only revert application state but also NOTIFY events published and queue jobs in flight. This alone makes Postgres a powerful option.
In fairness, Redis has only been available for 12 years so someone who has been in the industry longer has probably encountered systems using DBs as queues for no better reason than an alternative was not available when the system was made. (Rabbit just a couple years older I think.)
But in this day and age, you have better options to start with c'mon.
> they spent too much time fiddling with RabbitMQ, when Postgres would have been fine.
Rather ironic claim in a thread full of tips, tricks, and gotchas just to make PG behave like a queue no?
I can't speak for RabbitMQ personally but Redis has got to be among the simplest drop-in dependencies I've ever encountered. For a simple queue, the defaults do out-of-the-box.
Gut feel also says, Redis defaults for queuing would also serve more than PG defaults for queuing so you could leave it untouched for far longer to focus on your product, get to profitability, etc. Feel free to debunk with data.
I agree that there are many cases with low workload where that would be plenty.
But if you must use disk based table for Job queueing, set fillfactor = 50. This takes care of heavy updates.
Indexes are helpful but costs memory and CPU, so always make sure you partition the table based on job_type for performant pending job query.
I wouldn't recommend using LISTEN/NOTIFY unless you are okay with "at most once" semantics. I have used disk table based approach for PUB/SUB to replace Kafka. More fine tuned approach will also allow (job_type, consumer_group, publisher) as a partition key.
Ref - https://www.postgresql.org/docs/current/sql-createtable.html
I have been using RDMS all along since 1996.
Nokia NetAct was scaling GB of data across multiple clusters with OLAP reporting engine in 2005 with no hiccups.
The experience I had with DynamoDB kind of proved to me I haven't lost anything by staying in the SQL path.
Most NoSQL deployments I have seen, could have been easily done in Oracle or SQL Server, provided they actually had a DBA on the team.
It was a government project, written by one team (us) to be maintained by another.
The data that needed to be expunged was user signup data, upon completion the record was sent to a CRM and the Redis record destroyed. If the signup wasn't finished it's automatically removed after 12 hours.
Referential integrity wasn't really a problem, emails are unique and if we clash the two records are auto-merged by the CRM.
Setting up scheduled tasks, triggers, partitioning, cron, etc, is just more things that can go wrong. If they go wrong _and_ go unnoticed we end up with piles of data we shouldn't have. That would be many different kinds of bad.
Doing `redis.set(k, v, ex: 12.hours)` or whatever is just easier.
(I haven't used Kafka for a while, if anything below is outdated let me know)
The main issue I've experienced was balancing work across consumers - if one consumer is slower than others (e.g. running on a machine with some other heavy workload). In such case when a larger number of jobs is enqueued and expected to process quickly it's possible that everything will complete except jobs which landed on the partition belonging to the slow consumer. There can be only one consumer per partition so others can't help.
One could consider using a single partition to avoid this imbalance but this means that there could be only a single consumer - not good.
In other queue systems consumers fetch up to N jobs from a shared pool when they need more work. This means that in the case of a slow consumer the lag is limited to by the time required to process N jobs.
This situation also arises if consumers are equally fast but some kinds of jobs are considerably slower. Random distribution across partitions mitigates it to a degree.
My case was precomputing caches after some event, the goal was to do this within 10 seconds. First implementation with Kafka worked but very often some consumer couldn't use as much CPU time as others and we didn't want dedicated nodes for this because this even was happening at most few times a day. As a result jobs from a single partition were processed 2x slower - we had to overprovision.
Replacing Kafka with a queue based on Redis solved this right away, all consumers were finishing at the same moment. Although it wasn't a large scale operation.
The second complaint about using Kafka as a job queue is not being able to modify the queue, e.g. to reorder jobs, postpone jobs or cancel jobs
Direct visible effects are wrong decisions entangled in spaghetti-like complexity.
It's hard to reach technical bottleneck in well designed systems. Computers are really fast novadays. They will vary greatly depending on what kind of system it is. Out of resources – cpu, memory, network, disk io – likely the weakest of them will be saturated first – network. But that's not a rule, it's easy to have system which will saturate ie. CPU before.
In this example - great, you have a “queue” abstraction. When you switch to RabbitMQ or Postgres, how do you move your existing data without a quality of service or data loss? It can be difficult with production datastores even if the abstraction within your service is beautiful.
If you want a queue, get a queue.
If you want a relational database, get a relational database.
If you want a key-value store, get a key-value store.
Don't cross the streams.
A lot of people don't see the effects of their decisions. They leave a company after 3-5 years and go and work somewhere else where they get to make the same mistake again. The bottleneck indeed is lack of competence.
As for technical bottlenecks, it's quite easy to hit a wall. Be it through layers of stupid or unexpected success. We have unexpectedly reached the limit of what is possible with x86-64 on a couple of occasions due to stupid decisions made over 10 years previously for which there is now no longer the budget or attention to fix.
Sure, there are cases were you actually need strict ordering, but they are relatively few, in my opinion, involving things like transaction processing or other areas where the order of operations change the effect of the operations and must be strictly sequential.
Ordering itself brings significant technical challenges. For example: With a queue, you can only have strict ordering as long as you only process one item as a time, with no concurrency. Ordering also complicates handling of failures and retrying, because it means every newer item has to wait until the failed item has been dealt with; if you use things like dead letter queues, order is violated.
In almost all use cases -- anything from image processing to newsletter delivery -- a better system is to have an unordered list of items that anyone can take items out of, in any order. Then you throw as many workers at this system and design it so that there's minimal locking needed to prevent workers from stepping on each other's.
There are queue systems that have a purpose in a stack, specifically pub/sub brokers, which can be used to schedule work along several dimensions. You use the pub/sub system to signal work to be done, but each payload just refers to some state that's stored elsewhere.
It's a little insane that a highly rated thread on HN is telling people to use postgres as their queuing solution. The world is wide, I'm sure that somewhere out there there is a situation where using postgres of a queue makes sense, but in 99% of all cases, this is a terrible idea.
Also, SQS and nsq are simple.
As for triggers, reporting, audit, I'm laughing now because you miss the point. That's another bunch of IO in your monolithic black box, and you're still building it.
Start here: https://www.rabbitmq.com/prometheus.html
Now, if you’re using a nosql approach for data storage, then you already know your answer.
For a queue, you need a table with an insertion order and a status indicator. “Primary id” is not enough because “primary id” may not be ordered (e.g., a UUID primary key is a primary id.)
OTOH, while a DB can do this, therr are plenty of well-tested open source purpose built message queueing solutions, so the swt of scenarios where “roll your own with a DB” isn’t a misuse of effort is limited.
If you throw a proper messaging broker on the table like RabbitMQ then you're getting message acknowledgement as well which is important if you must do the work. A failed consumer will result in the work being rescheduled on another consumer transparently. Of course there are metrics you can see and get heads up when your consumers are not working.
Ergo, don't simplify something too soon. Even reliable delivery of newsletters is rather important otherwise why would all the tracking pixels exist? (they are a form of message acknowledgement)
The point of queues in any particular application is some subset of that; the calculus of implementation approaches that make sense depends not on the abstract point of queues in general but on the concrete point of queues in your use case.
They'll also manage the consumers of the queue and scale them too! Serverless is bliss.
To clarify we just moved this entire problem to SQS.
In the Drupal case, the database returns a single row which is pretty much guaranteed to be the same for all workers in between updates. You really don't want that in the same database your page views hit. At least selecting N rows at a time and then claiming a random item from them would be slightly better.
All to save 2 to 3 hr of Googling for the best queue for your use case and finding s library for your language.
It makes sense if you don't care about reliability and just need something easy for many many people to deploy (ex: Drupal).
Of course, if scalability is not a concern (or the connection limitations are eventually fixed in postgres), this would be a very viable approach.
Brandur wrote a great article about that here: https://brandur.org/postgres-queues
If you need to keep track of old data, then yes, migration is hard. But queues don't save old data—that's the point of a queue.
By using SELECT ... FOR UPDATE SKIP LOCKED, the record will automatically get unlocked if the worker crashes.
> It's hard to reach technical bottleneck in well designed systems. Computers are really fast novadays.
I have been listening to how fast moderen computers are for the better part of the past two decades, yet as I user I still have to deal daily with too many of slow software and slow web services.
If we're talking PostgreSQL specifically, and newer-ish Postgres (9.5+ I think), then you can leverage its abilities to do all this in one atomic query:
UPDATE jobs
SET status='working'
WHERE id = (SELECT id
FROM jobs
WHERE status=NULL
LIMIT 1
FOR UPDATE
SKIP LOCKED)
RETURNING idRedis is really a mix-bag of many useful commands, see [https://redis.io/commands]. It has been referred to as the "swiss army knife" because you can build many custom solutions from these building blocks. Still, its most common use-case is for in-memory key/value caching for performance reasons. For instance, if you have one database query that takes a particularly long time, you can execute it once and store the result in redis, and then retrieve that value extremely fast.
https://wakatime.com/blog/45-using-a-diskbased-redis-clone-t...
Mongo said “Throw data at me and I’ll scale with very little work”.
Now, I’ve always largely believed that’s penny wise and pound foolish but it’s certainly a good pitch.
UNLOGGED
If specified, the table is created as an unlogged table. Data written to unlogged tables is not written to the write-ahead log (see Chapter 29), which makes them considerably faster than ordinary tables. However, they are not crash-safe: an unlogged table is automatically truncated after a crash or unclean shutdown. The contents of an unlogged table are also not replicated to standby servers. Any indexes created on an unlogged table are automatically unlogged as well.
> All to save 2 to 3 hr of Googling for the best queue for your use case and finding s library for your language.
The cost of using a new piece of tech in production is not just 2 or 3 hours.
Saying this as someone who has been on both sides of that interview question, and also evaluated performance in it as a hiring manager.
If your load requirements fit using your database as a queue, it can radically simplify your system.
I've personally done a fair bit of migration from formal queue systems into an RDBMS. You simply get a lot more control and visibility that way. But I use task queues heavily too. It's all very application-specific and I don't think you can generalize.
Absolutely and that reason is, you are still a small team, with a small user base to boot. That's fantastic opportunity to learn a new technology and build on it properly! Remember everything is easier in software engineering if you assume you have no users[1] and your situation is as close as it gets to this ideal. Leverage it.
Plus, as me and others keep saying, Redis (and other proper queues) isn't a complex addition to your infra. This isn't Hadoop, or Kafka, which is a very special type of queue (one way to put it, at least).
> one request per second, and postgres for a queue is more than sufficient for that
Yes I agree but...
> Or would we do just as well to abstract away the queue in the code so that we can switch to redis when we do run into scaling problems?
What I read when I see such statements is this mythical software engineering ideal that with enough abstraction, a migration is just a matter of writing a new class that implements some interface and then changing a config. For a sufficiently complex app infra, that happens almost never because you could never keep the abstraction leaks to an acceptable level.
Another thing, abstraction does not solve all your problems if the underlying implementation is poor fit to begin with. Let me paint you a plausible scenario:
Once you are large enough, you might find your PGQ acting weird and you realize it's because someone in the team wrote code that accesses your queue table like it's an actual table of records, not a queue. So you think, okay let's prevent that from happening. Maybe you add users and permissions to distinguish connections that need to access between tables proper and queue. Maybe you start writing stored procs to check and enforce queue invariants periodically.
Well, guess what, all those problems would've been solved for free if you invested maybe one work day getting a Redis server running when you were a two-person op serving one request per second.
Lastly, scaling a relational DB is an entirely different beast from scaling a queue. Scaling anything does not ever come painless but you can reduce the suffering when it comes. Would you rather scale PG so it can keep acting as a queue or scale a queue that's, you know, really a queue in the first place? Heck the latter might even be solvable by throwing money at the problem (i.e., give it more compute).
[1] Except for the part where you need to make money, of course.
At $JOB-1 I wrote a `Queue` and `QueueWorker` abstraction that used an environment variable to switch between different queue backends while providing the same interface. Because of this I got everyone up and running with Redis lists as a queue backend and then could add in things like MQTT or RabbitMQ as things evolved. I also defined very loose constraints for the queue interface that made it so the implementer. Essentially there was a `push()` which added something into the queue or failed and returned an error. Then there was an `onWork()` which was called whenever there was work "at least once" meaning your system had to handle multiple instances being delivered the same work item. We would only ack the queue message after `onWork()` completed successfully.
There's not really anything preventing a team from doing this, putting a pin into it, and coming back to it when there's a scalability or reliability concern.
I don't think adding another component (especially one with storage/backup requirements/more complex failover procedures) should be taken lightly.
Paying attention to the tips/tricks/gotchas is something you'd pay once and hopefully document in source code/documentation.
If you piggyback off someone else's work (another team at your job, a cloud-based solution that handles everything, etc) the calculus would change, of course.
Case in point our team went with Redis, just the default, use it blindly without fully understanding our requirements and how redis helps scale.
2 years later we spent 2 sprints, holding back the release trying to understand RDF vs AOF, and why we're seeing massive spikes in consumption and performance triggering pod eviction, running comparison tests to prove which works better and explaining why, running qa tests (regression, performance, load), introducing postgres for queuing, redoing our code to bypass the sync mechanism between how data flows between redis and postgres, updating dependenciies, migrating existing customer data (various on site locations), explaining this to team members, managers and their managers, installation technicians, support engineers and presenting it at engineering townhalls as a case of bad decisions.
Not worth it.
How sensitive is your app to latency? How much data and request volume you need to handle?
Do proof of concepts, write thorough load tests and go for what makes sense.
Either way, no matter which tech you choose, make sure monitoring and alarms are in place and that you do regular maintenance exercises.
3. When processing is complete the matching table entry is deleted.
4. There is a "reaper" process that checks for stale jobs ( based on time in the future ) and resets them to 0.
This of ouffcourse raises more questions.
Do you really need PostgreSQL? How to Get Away with Just SQLite
Exclusive ownership of a SQLite database by the application can deliver better performance than exclusive ownership over a single node postgres database. This also extends to SQL Server, DB2, Oracle, or any other hosted solution.
Also reading your reply I get the impression that "sync mechanism between redis and postgress" was the bottleneck. Wondering if you can add some details around it and also was this something that can't be fixed by fine tuning redis config, rather than completely removing it from your stack.
There were many problems but at the core of it, this was us having redis write huge amounts of data to disk very frequently causing this.
We could not reduce the frequency (product would not allow) and we couldn't find a way to make the writes reliably fast.
I like to believe there exists a possible way of handling this, but point being, our team had no time to find out, how redis works internally and have confidence that the new way won't bring up new surprises.
There are libraries in many languages written specifically to accomplish the "make PG behave like a queue" part. All these people are explaining the principle of how it would work, but in practice, you can just pull in one of those libraries and be done. (Or, more accurately, you use an abstract external-job-queuing library and tell it to use its Postgres adapter, which in turn uses that library.)
In exchange, you get the ability to not have to set up / maintain any more infra than you already have. Which can be important if you're already "at scale", where every component you set up needs to be duplicated per geographic-region, clustered for High Availability, etc. If you've already done all that for Postgres, it'd be nice to not also have to do all that again, differently for Redis, RabbitMQ/Kafka/NATS, etc.
> Gut feel also says, Redis defaults for queuing would also serve more than PG defaults for queuing
Don't know about what you've got your queues doing, but our service only does (persistent, out-of-process) queuing for two purposes:
1. Global scheduled background-job queuing. (Think "cron jobs.") There are only a static number of these; they don't scale up with load.
2. Lifecycle email dispatch. This scales up with MAU, but with very very low coefficients.
For those cases, Postgres would totally be enough. (Anything would be enough!)
Most "queues" needed in the real world are requirements-light like this. It's a rare company whose product is doing something Apache BEAM-alike with its data, where a high-throughput reliable MQ is a fundamental part of the data workflow. (And those companies, I hope, know that they need an actual MQ, not a piece of software that does queuing in its spare time.)
Our company doesn't use Postgres queuing; we do, in fact, use Redis for it instead. But we only ended up doing that, because we already needed Redis for other stuff; and if you already have Redis in play (including an ops plan for scaling it), then it's the better tool for the job.
Thing is, because purchasing is often shared across third-parties, even then you can’t guarantee that you’ll have the reservation until it’s paid and entered into the shared booking backend… unless the third party backend has a locking mechanism, at which point you’re probably not using your own Postgres to do this, but their mainframe instead.
Other things that matter, often more than what's easiest:
* Flexibility. If you aren't 100% sure what you'll need, but need something in place today, it can make sense to choose an unopinionated, generic solution.
* Familiarity. Maybe you've never used Rabbit.
* Ops simplicity. Maybe you already have a bunch of postgres deployed. Adding a new type of system means you need new ways of monitoring, new dependencies, different deployment.
That's... nothing. Databases handle billions of rows effortlessly using a b-tree index. So not really sure what point you're trying to make?
One would need the appropriate transaction isolation, but shouldn't it work?
It's also probably one of the easiest services to deploy and manage; often a one-liner.
Plus like you said, swiss army knife. It has so many uses. It's inevitable my stack will include a redis at some point, and my reaction is almost always "I should have just started with redis in the first place."
Is redis prone to golden hammer syndrome? Of course. But as long as you aren't too ridiculous, I've found you can stretch it pretty far.
Databases scale well in general, it's one of the things they're built for.
A single database server with an indexed table (b-tree) and an SSD is extremely performant. (And yes there are other indexes that could be even more performant, but b-trees are already extremely fast.)
But the huge advantage of putting a queue in your database is being able to create transactions across both the queue and data in other tables that needs to change at the same time.
Not to mention one less extra tool to maintain, existing database tools for backups replication etc. automatically include the queue as well, etc.
As well as the fact that it's incredibly easy to make your queue behave according to whatever kind of custom business logic you might need. (Which is where additional indexes and b-trees can turn out to be necessary, and which is precisely what databases are great at.)
So to the contrary -- unless you're at Facebook-level scale, using modern databases for queues generally seems like an excellent choice. (And in the special cases where it's not, that's usually quite obvious.)
Don't be scared of having to make changes in the future. Do the small amount of work it takes today to make sure your transition in the future is easy.
Transitioning from a SQL queue to redis it's only difficult if you have a bunch of SQL throughout your code. If you have that, you did it wrong.
It might seem like everyone works for a startup, but big companies have a lot of engineers, and we still have to solve new problems there even though the company is not new.
But I've run billions of queries per day against a single Redis instance with zero failures serving up traffic to large, enterprise-level customers with no downtime and no major issues (knock on wood). The only minor issues we've run into were some high-concurrency writes that caused save events and primaries to failover to replicas, and resulted in a few minutes downtime at the beginning of our experiments with Redis-first approaches, but that was easily mitigated once we realized what was happening and we haven't had a problem since.
Huge fan of a Redis-first approach, and while the haters have _some_ valid points, I think they're overstated and are missing out on a cool way to solve problems with this piece of tech.
So while this might be good advice if you will need to scale, it's certainly not blanket advice.
In the first case, I would much rather just have Postgres and not a separate queue to manage.
"Do you really need SQLite? How to get away with a dozen secretaries, pencil, paper, and a fax machine"
I'm with ya. Why not use the right tool for the right job, given the right tool exists?
Also keen to know if you saw any disadvantages with this approach?
Yes, exactly.
> This sounds like it might be vulnerable to dead consumers causing WAL to pile up without the right settings/periodic checks.
There are some subtleties around this indeed. You should monitor backlog of replication slots, so to identify inactive consumers as you say. Also, there are some corner cases you need to take care of: when listening to changes from a low-traffic database on the same PG host that also has a high-traffic database, the replication slot may not be acknowledged often enough. Debezium mitigates this by optionally writing changes to a dummy heartbeat table in the low-traffic database, so to advance the replication slot.
In Postgres 13 there's also a new option "max_slot_wal_keep_size" which limits WAL size retained by a replication slot. This prevents unbounded WAL growth, at the risk of consumers to miss events if they are down for too long.
All in all, proper monitoring and alerting is key.
Avoid consuming that queue directly though -- which is probably what you're thinking when saying this is a dumb idea and I tend to agree. Typically, you want to have a worker that loads entries into a more optimal queue for your application.
Bottom line though, there is not a single best "queue" product. There are lots of queues offering wildly different semantics that directly impact use cases.
> committed atomically
Complex billing systems don't work that way. Worker processes are not always in these neat boxes of "done" or "not done". Much like rollbacks are a developer myth. If a process was that trivial then you wouldn't need a queue and workers in the first place!
> Also I'd detect a worker has died by recording the start-time and using a timeout.
There are many ways to solve this and many ways to get it wrong. Not working in UTC? Oops, better hope nothing runs during daylight savings changeover. Parent process died but worker finished job? Let's hope the parent isn't responsible for updating the job completion status. Large job is borderline on the timeout? Better hope parent process doesn't restart the job while the worker is still working on it. Network partition? Ut oh. CAP theorem says you're out of luck there (and typically there is at least one network hop between the DB server/controlling process and the system running the workers).
Probably the more straightforward solution is to give each worker an ID and let them update the database with the job they pick up. Then, something robust like systemd, would monitor and restart workers if they fail. When a worker starts, they find any jobs where table.worker_id = myId and then start back on those. But you still have network partitions to worry about. Again, not at all trivial.
One can only have bought into Mongo's story, by lacking the skills to understand how fake it was.
Celery doesn't use Redis under the hood.
However, you may use Redis as a message broker and/or result store.
https://docs.celeryproject.org/en/stable/getting-started/bac...
In the current reality today, implementing everything in Postgres is probably going to be slower to market (i.e. for a start-up) than using off-the-shelf products. When you do need to scale, this is when you get to learn about how valid your assumptions were in your abstraction layer - mostly likely in production. As a concrete example, Kafka isn't designed to work well with large numbers of topics. Similarly, InfluxDB isn't designed to work well with high cardinality time series. I think it is generally wiser to "skate where the puck is going" in this situation.
Of course, everything is a trade-off. Postgres is incredibly reliable (like insane) and simple to operate. I'd say for any kind of internal line-of-business type application where scalability is less of a concern you really would be doing your ops team a service by implementing everything in Postgres.
With those new tools you had other companies building on top of those databases for far cheaper than a license to MSSQL or any other OLAP of choice would give you.
As long as you don't need clustering (even a single node can handle some pretty heavy load), it's actually really simple to setup and use - way easier than Postgres itself, for example.
My biggest beefs with it have historically been the Erlang-style config files (which are basically unreadable), and the ridiculous spiel of error messages you get if you have an invalid configuration. But thankfully RabbitMQ switched to a much simpler config file format one or two years back, and I understand the Erlang OTP is working on better error messages and stack traces.
And, you mean, a PGQ will not need any special monitoring other than the common DB metrics?
For instance, if I ever run a queue, it's just due ops diligence to know the length of the queue, average time an item spends in queue, throughput over time, among others. Are there standard monitoring modules that would check this for a PGQ? Because in exchange for setting up a proper queue cluster, compatibility and documentation for common stuff like this is also what you get.
The first one is particularly a sticking point for me. If you don't do it right, you will end up issuing a COUNT on a table periodically. You might say it's acceptable for most companies, because they don't need high-performance queues, but I specifically need my monitoring to be reliable during times of unexpected high load. Also, ideally, there is close to zero chance that my telemetry is the one that ends up causing me trouble.
Also, why does no one consider the added "surface for human error" that using a DB for a queue introduces. For instance anyone can drop in the DB and perform a query you were not supposed to. A malicious actor can update all outgoing emails in the queue to another recipient. If these are sensitive emails like password reset or OTP, good luck. A dedicated queue process does not allow such operations.
I have to say, it's very perplexing to me that people (especially in HN) would avoid using the right structure just because they have to invest a bit more work.
> use it blindly without fully understanding our requirements and how redis helps scale
I'm sorry I don't get how I could come across as advocating the use of Redis blindly. My point is if your data flow looks like a queue, then use a queue, don't hack a relational DB to become a queue. I think that's reasonable rule of the thumb, not going in blind.
You're imagining this as if it's just the CapEx (the work of setting up Redis), but in reality it's the OpEx (the work of maintaining a Redis cluster) that kills you. Think of it in terms of the number of ops salaries you have to pay. At scale, this number goes up in proportion to the complexity of your stack.
I would argue that if you've built a system up from scratch, this is much easier to debug and maintain than a foreign piece of software. Rabbit is just way overkill if you have a couple of jobs per hour, for example.
Repeating my other comment: do you mean to say using an in-DB queue will not need special monitoring other than what's already standard metrics for DBs?
Since a queue is a different structure, I would want to know metrics that might be unfamiliar for standard DB monitoring. Queue length, queue throughput, etc. Can I get that from my DB queue with negligible cost?
Another thing, there might be standard practices and patterns for in-DB queues, often encapsulated in a nice library/middleware. But as databases are not queues there is no assurance from the vendor that those patterns will remain applicable in future versions.
> Paying attention to the tips/tricks/gotchas is something you'd pay once
Actually, I disagree with this. Everytime someone connects to that database, they have to remember that one special table which is used as a queue, and remember to tread differently where it is concerned.
Setting up a queue? You do that once and everyone who interacts with it will not forget it is a queue. Heck you can't even do your usual DB operations there. How can you not remember it is a queue?
That being said, I agree with other comments that this is somewhat swimming against the tide. Redis is much more commonly used, and so if you don’t mind adding another service into the mix, I’d probably recommend going with Redis instead.
I've had so many horrible experiences with DB-based queuing over my career once you get to a certain scale. Just use a message bus, bro!
Using an in-DB queue doesn't give you zero OpEx does it? Maybe you can make the argument that it's cheaper but it's cheaper for a reason: the job wasn't done properly, so to speak.
Both options introduce new failure modes and op costs into your system. Might as well do it properly if (and this is a huge IF in my opinion) for slightly more cost. When you run into a failure it's standard, maybe the solution is even one Google search away as opposed to realizing, one hour into a late-night debugging session that, shit, that table should be treated as a queue!
The database records the start time, in the databases system time, in UTC.
The parent process can't update the job status, it can only cancel jobs.
There's a pool of workers. Each working may fetch a pending request or wait. Once it has fetched a request, which sets it as pending, no other worker is able to fetch this request. Only this worker may alter the status of its request and it may only set it to completed, and it may only do so with its results.
The worker periodically writes its status to a separate table. If such status has not been written in a timely manner, the worker is assumed to be stalled and a new child request is created, available to other workers.
Putting this logic into the database instead of a parent process, which managed a shared concurrent queue, has been a by far more pleasant solution. Granted, we're not talking large numbers of concurrent requests. Ten as opposed to ten thousand. If we're talking hundreds of concurrent, short requests, I can see this starting to become problematic.
I think there's huge difference in the needs of different implementations of queues and workers so I'm interested to hear more about workloads where this approach fails and specifically in which regards. In my experience, it's been pleasant.
I've seen this twice in this thread, but I don't know what that means. Can you explain a bit?
I have been thinking about using the library, mainly because it's less operational hassle than running another service that has to be sticky to one host with persistent storage and backups.
MS also moved from it afterwards when they acquired Skype but I guess it was more because of the shift to SQL Server.
Re: Monitoring
Presuming you're already monitoring anything at all, adding monitoring for a new piece of infrastructure is a pure CapEx cost. You presumably already have monitoring infrastructure running. Its ingestion, by design, will be 99% idle — i.e. it won't need to be scaled horizontally proportionally to the number of components. The only thing needed, then, will be careful up-front design for the monitoring.
Which is usually also a cost already paid for you in advance when you use an abstract job-queue library. They all know how to expose Prometheus metrics endpoints, and they do the accounting to serve those endpoints efficiently (usually using process-local per-job-producer and per-job-consumer counters, which you must then roll up yourself at the PromQL level, taking the irate() to find spikes.)
Re: AAA
> Also, why does no one consider the added "surface for human error" that using a DB for a queue introduces.
Because the type of person who reaches for their DB first to solve a novel problem, is the same type of person who understands and makes use of their DB's security features, to the point that doing more DB-security-config isn't a complex one-off problem for them, but something they can whip off in a few minutes.
For a simple option, you can create two service-account DB roles: a queue_producer, and a queue_consumer. You can put all the job-queue implementation tables in a schema owned by the queue_producer; and then grant the queue_consumer SELECT privileges on all the tables, and UPDATE privileges on some of them. Then, nobody but the job_producer (or a DB superuser) can create or destroy jobs; and nobody but a job_producer, a job_consumer, or a superuser, can read or modify jobs. (Your job-queue abstraction library within your service usually maintains its own DB connection pool anyway, so it's no sweat to have those connections use their own credentials specific to their job-queue role.)
For a more complex — but perhaps useful? — option, the tables themselves can be "hidden" behind stored procedures (DDLed into existence by the abstract job-queue library), where nobody has any privileges (incl. SELECT) on the tables, only EXECUTE rights on the sprocs. And the sprocs are carefully designed to never do anything that could have an unbounded CPU time. Then anyone can "check up on" or even "insert into" the job-queue, but nobody can do anything "funny" to it. (My god, it's an encapsulated API!)
Once again — the libraries that abstract this away, already think about concerns like this, and choose one or the other of these options. That's why libraries like this exist, when the "core premise" is so simple: it's so there's a place to put all the fine details derived from man-years of thought on how to make this approach robust.
-----
On a completely different track, though: having a queue in the DB can sometimes be the optimal (i.e. the "if we had infinite man-hours for design + implementation") engineering decision. This case comes when the thing the queue is operating upon is data in the DB. In such a case, the DB data modification, and the job's completion, can succeed or fail together atomically, as part of a single DB transaction.
To accomplish the same thing when your queue lives outside the DB, you usually end up either needing some really gnarly distributed-locking logic that both your DB and your app layer need to know everything about (leaky abstraction!); or you need to completely remodel your data and your job queue into an event-streaming paradigm, so that you can "rewind" one side when the other side fails.
Hopefully the emphasis is clear why this is silly.
https://mikehadlow.blogspot.com/2012/04/database-as-queue-an...
In these days with Docker, ready cloud solutions getting the message broker is so simple that there is no reason to use database as a broker. If we know that our traffic will be so small, just install DB and message broker on the same VM, to avoid spending more on hardware or hosting.
With redis clustering, there's no guarantee the data has been replicated. I'm not even sure there's any guarantee the data you just asked to be recorded be stored even once if a power outage happens immediately after the request.
If you're actually trying to solve the problems Redis is meant to solve, just use Redis. Otherwise you become the sole maintainer of a glued together, suboptimal version of Redis based in Postgres. For most people, they'd rather build whatever caused them to need Redis in the first place, rather than building their own internal worse-version-of Redis.
They allow overbooking the flight but you can't book the same seat twice.
Overbooking can make financial sense -- even if you need to hand a little compensation from time to time it's still better than flying with empty seats. Of course it sucks big time for the travelers especially if there are no volunteers. IDB is miserable. https://www.transportation.gov/individuals/aviation-consumer...
The things that often make databases appear slow is that they’re typically configured for large random reads rather than small uniform writes. And they have insane initial response times because of all the IPC they typically do when starting a session.
But aside from that, if you use a heap table without an index you’ll can probably get similar throughout to Kafka. And if you turn Fsync off you might even get up to REDIS speeds.
I have to disagree. Of course code quality is important, but building things "properly" because "we may need it later" is a great way to kill a project with complexity. KISS, YAGNI. An early startup is, IMHO, not a good place to learn about new frameworks while getting paid - you're on borrowed time.
Make a back-of-the-envelope calculation about how much throughput you need. E.g., if you expect to have 10,000 users, and each may make one request per hour, you're dealing with 3 qps. Anybody who wants to bring in a new dependency for this, needs some talking to.
(If you already need Redis anyway and it's a better fit than Postgresql, then sure, go ahead.)
There’s a reason SQL is still around after all these years. It’s basically math and while the syntax is a little moldy the principles are eternal.
Yes service accounts are fine and dandy but [a] isn't that more hoops borne by choosing the wrong tool for the job (gotta stay mindful of those permissions, you can have an Ansible typo, etc) and [b] as I said, proper queue services would not even allow tampering of enqueued data. For no extra cost other than installing the service. In contrast, your service account falling into a malicious actor is still a threat vector.
> For a more complex — but perhaps useful? — option ...
Now this is just a queue service with extra steps running in a relational DB instead of natively as an OS process. You did cite it as just an option but I don't see why this is an attractive option.
> I didn't see any mention of anything I'd consider an OpEx cost.
It seems to me we have differing definitions of OpEx and, judging by your previous comment, you value this almost exclusively in terms of "the number of ops salaries you have to pay". Even if I play by that definition, I could tell from experience (and also corroborated by other commenters here, in other subthreads of this discussion), that operating a Redis cluster does not need extra warm bodies. The people who monitor that your app responds, your servers aren't running out of memory, could take on the task with little to no additional training.
The experience I want to cite, bearing in mind of course that everyone's mileage varies: in my previous job, customer base of 2M and growing across two countries, we operated and monitored a redundant Redis cluster with an ops team of 3. In my first job I talked about in another subthread here, we managed three high-traffic Redis queues (and a few other smaller-traffic special cases) with a team of 7. In both jobs, we also did active across-the-stack development; we weren't dedicated to monitoring, nor to the components related to the queues.
In fact I would argue running a DB (queue or no queue) is just more complex than running a queue service. Rather telling is the need for "DB Administrators/Specialists" but no such equivalent for Redis or other proper queueing solutions.
> Which is usually also a cost already paid for you in advance when you use an abstract job-queue library. They all know how to expose Prometheus metrics endpoints...
Honestly, this sounds fine but I'm rather wary of the caveats that might be present. How do they do it efficiently? What do I need to keep in mind to keep things efficient? As such, unless you can point me to your preferred job-queue library which does all these wonders, I have to reserve judgment.
Edit: I just remembered mentioning this in another subthread. Even if you have a fantastic library/middleware abstracting all those queue concerns, that ties you in with that library. If someone wants to build integrations into your queue, they have to play by that library. If majority of your app is in Java (and so is the library), and Data Science (who, in this example, uses Python, not improbable to happen) wants to share your data pipeline, if that library isn't available for them, tough luck I guess? More dev time for someone.
And also, whatever underlying DB features or semantics your library might rely on in order to enforce a queue structure, you can't be assured that future versions of your DB would support those features because, you know, your DB isn't really in the business of queues. It opens you up to running an outdated DB version just to keep that queue functionality.
If you don't need clustering, RabbitMQ really is pretty simple to setup.
The common workaround is to use a connection pooler like PGBouncer so that clients reuse connections. This approach doesn’t work for LISTEN because typically a client will listen for its entire lifecycle so you can’t share connections in a pool.
That is, if you have the computational resources to service your entire queue within the required latency, then ordering doesn't matter, because workers could simply claim the first available random task, and it doesn't matter who finishes first or last. If you don't have the resources, then ordering still doesn't matter, because you are blowing your latency budget anyway.
For example, say users are uploading photos which go into a queue for scaling, cropping, etc. Let's say 10 users per second are uploading a photo. In the best-case scenario, the queue is empty, so all users should see their photo fully uploaded quickly. Worst-case scenario, your queue has a million tasks pending, which benefits a handful of early users who see their photos completed, whereas most users will simply wait forever. Prioritizing early users (i.e. queue order) doesn't benefit anyone when everyone will be miserable, so it's an arbitrary choice.
The ordering imposed by queues can create an illusion of fairness when task execution time (i.e. end-to-end latency) is not taken into account. If one user uploads a 1 gigapixel photo, it's most likely going to take up more processing time than other users, and again, there's no reason to prioritize that user over any other user just because their request arrived earlier.
If you have random-order delivery, you can still measure latency, capacity, and utilization, as well as have retries and acking, and utilization-based scaling.
Note the tradeoff doesn't make sense as soon as you're operating at a meaningful scale. A small likelihood of failure at small scale translates to "I expect a failure a million years from now", whereas at large scale it's more like "a month from now". Accepting the same percent risk of data loss in the former case might be OK, but in the latter case is irresponsible. Provided whatever you're storing is not transient data.
I've thought about using rabbit mq, but have a few questions: - it appears that when a job is consumed, it's gone. How would I maintain a record of the jobs & status of the job in rabbitmq? If I wanted to display job updates how would I handle that? I didn't think you could update a message once it's already in the queue
Or am I missing the mark? Do I want to separate the business entity "job" that maintains a status and updates and such from the "thing that dispatches jobs to workers"? And when a worker Has the job it just updates the business entity in the database?
It’s a killer.
If it’s “just a cache” OK. Fine. If you need it for something important like queuing or certain caching scenarios - be aware it won’t scale well.
It's always about tradeoffs in context, but I have seen plenty of instances of someone setting up a rabbit cluster for something that could be done trivially in their existing db cluster via the above.
Emphasis added because I feel like I addressed this in the paragraph immediately after the one you quoted:
> Plus, as me and others keep saying, Redis (and other proper queues) isn't a complex addition to your infra
I'm speaking out of experience and, as I already pointed out in another subthread, Postgres is far more complex than Redis. Consider the presence of "DB Admins/Specialists" and the lack of counterpart thereof for Redis and other queuing solutions.
Of course, if queues are not central to how your platform operates, you might be able to get away with Postgres. I still advise using Redis as a reasonable hedge against someone famous tweeting organically about your service because in this case, you don't want your DB to go down because some queue table had a surplus of transactions (or vice versa).
Not to mention, at an early stage, your tech decisions set precedents for the team. Maybe you have 10K users with a low qps but soon you are sending marketing emails to them and your system has periodic bursts of queue activity for all 10K users at once. When discussing this marketing "feature" rarely anyone thinks, "Hey we can't do that with our Postgres queue", rather "Yeah I saw functions in our codebase for queuing---this is doable". This is a small effort but a huge technical investment for later on.
Specifically I similarly pull rows into a 'queue' like that linked sql by WHERE primary key, primary key, and queue_status ORDER BY timestamp, uuid LIMIT 10. I was worried what would happen with many requests around the same time if it would accidentally pull duplicates into 'queue.'
I think hopefully I've implemented Spanner locking read write transaction correctly and that this won't cause problems at scale (we're not even close to google scale lol).
Your DB then shares a WAL log with your queue. Meaning a single managed physical replication pipeline for them both. Meaning only one set of leader-election issues to debug, not two. Meaning one canonical way to do geographic high-latency async replication. Meaning disaster recovery brings back a whole-system consistent snapshot state. Etc.
Honestly, if I had my way, every stateful component in the stack would all share a single WAL log. That’s what FoundationDB and the like get you.
> In fact I would argue running a DB (queue or no queue) is just more complex than running a queue service.
Well, yeah, but you usually need a DB. So, if you’re going to be paying the OpEx costs of the DB either way, then you may as well understand it deeply in order to wring the most productive use you can out of each OpEx dollar/man-hour spent.
(I feel the same way about Redis, as it happens: if you need it, and are locking your code into its model anyway, then you may as well take advantage of its more arcane features, like Redis Streams, Lua scripting, etc.)
However, maybe our company is uncommon in how much our service literally is doing fancy complex DB queries that use tons of DB features. We’re a data analytics company. Even the frontend people know arcane SQL here :)
> that ties you in with that library
The difference between what you / apps / abstracting libraries do in Redis, and what they do in an SQL DB, is that in the DB, the shape of everything has to be explained in a vendor-neutral manner: SQL DDL.
Sometimes Redis-based solutions converge on conventional schemas; see e.g. Sidekiq’s informal schema, which several other queuing systems are implemented in terms of. But when they don’t, there’s nothing you can really do — beyond hacking on the libraries involved — to bring them into sync.
In an SQL DB, anything can be adapted into the expected shape of anything else, by defining SQL views. (Heck, in an SQL DB with Redis support, like Postgres with redis_fdw, the Redis data can be adapted into any shape you like using SQL views.)
And that’s further enabled by the fact that the DB had received from the app, through DDL, a schema, that you can examine, manipulate, and refactor; or even synthesize together with other schemas.
> you can't be assured that future versions of your DB would support those features
You can if those features are in the SQL standard. I’ve never heard of a DBMS regressing on its level of SQL standard support.
- You have a separate schema for your procs
- You define your procs in files
- You write tests for your procs
- You put your files into git
- Then, in a transaction, you drop your old schema and deploy the new one
If we rephrase question to the material reality of "What will we do when we're pulling in $20,000,000 a month?" then you can see how silly it is.
I dunno, just buy a company then and use their stuff? It's literally not a problem that a pre/early-revenue company should or realistically can solve. It's part of "founders delusion" - the same kind that make them get a huge office that's mostly empty or all that unused computer hardware sitting in the closet. It's just gambling.
It'd be like saying "we have $500,000 in the bank at this new restaurant and need to be ready service 5,000 national locations with our staff of 5."
I mean no, by the time that's no longer just a mental exercise (as in you have, say 4,700 locations) you'll be a different kind of institution better able to solve the problem. Literally just forget about it. It's so profoundly irrelevant...
Solve the problems that are actually real, you've got enough of them, I promise.
1. Recovery.
2. Concurrency.
3. Polling.
In more detail:
1. Recovery. Suppose a worker dies while processing a job. The job should be retried, but how do you know this, since expire > 0? You can impose a timeout, but that has drawbacks -- there might be long jobs you repeatedly start but never finish. To recover failed jobs without imposing a timeout, you'd have to run both the UPDATE and the job processing inside a transaction. That way, a failed job can result in an implicit ROLLBACK, freeing the job for future worker attempts.
2. Concurrency. So now recoverable workers are trying to claim jobs inside a transaction. If there are 10 jobs and 10 workers, we want all 10 jobs processing concurrently. However, each worker SELECTs the first item_id and tries to UPDATE that same row. 1 worker wins. The other 9 workers block while that transaction completes, then update zero rows. Concurrency will hover around 1.
3. Polling. In the transactional UPDATE approach, there's no way to tell whether there are free jobs short of trying to claim them via UPDATE, which blocks. So you must poll, and either burn cpu on quiet queues, or introduce an artificial delay into job processing times. The beauty of the SKIP LOCKED approach is that it can tell whether there are free (not currently processing) jobs. Even if you wake all 10 workers up via LISTEN / NOTIFY for a lone job, 9 workers in the thundering herd will fail to claim the job, and go back to sleep.
I blame TFA for not sufficiently explaining the SKIP LOCKED approach. A much better explanation is here: https://www.2ndquadrant.com/en/blog/what-is-select-skip-lock...
My general recommendation is to engineer for only 1 order of magnitude greater than you need today, but even that stops applying eventually.
I feel like the understanding we can come to here is that we have differing definitions of necessary complexity.
You did mention you work in data analytics and I have worked and am working in more traditional product-oriented gigs. Everything you mentioned are nice, impressive even, but to me they are toys. I don't need a queue with a WAL stream, or PITR backups. Queues, to me, are inter-process communication mechanisms, or a means to concurrency. In fact, worst case, you can delete the queues or restart Redis even without triggering a disk write (though I note this is another feature that comes free with Redis); it would inconvenience our customers but they can always just retry.
Of all the benefits you mentioned, leader-election is the only one I could make a good case for.
> then you may as well understand it deeply in order to wring the most productive use you can out of each OpEx dollar/man-hour spent.
Understanding it is one thing but using that understanding to create/develop (CapEx) and then maintain (OpEx) an in-DB queue feels like unnecessary costs. CapEx to install Redis is practically nil, and you seem to agree that running it is far simpler OpEx than PG too ("Well, yeah, but...").
Of course, I keep in mind your point about libraries doing the dirty work. But if it all boils down to a third party in the end, the point I'm replying to above is rather moot no? And if it comes to a third dependency anyway, I've already reasoned out my preference for an actual queue solution.
> explained in a vendor-neutral manner: SQL DDL
> You can if those features are in the SQL standard. I’ve never heard of a DBMS regressing on its level of SQL standard support.
While this is an ideal case, my experience so far shows that every major DB package relies on their own tweaks to the SQL standard, and so "If those features are in the SQL standard" turns out to be a pretty big "if" to ask. I don't worry about vendors regressing on SQL standard support but rather that the libraries for DB queues are relying on the non-SQL-standard features/behavior to mock out a full queue functionality---non standard behavior that has no guarantee of consistency across versions.
I mean, if there's a DB queue middleware/library that works across DB vendors, be it Postgres, Oracle, Maria, MySQL (with a reasonable storage engine, perhaps) then that's a convincing argument that the SQL standard can support queuing sufficiently. But otherwise, this counterpoint is a hard sell for me.
Security and PI stuff is unfortunately something you do have to be super anal about these days but scalability? no.
How you setup LISTEN and NOTIFY is app dependent. In a multi-tenant database, you could have 1 NOTIFY channel per tenant.
As you scale, you probably do something that listens in a smarter way, maybe 1 Go channel per client with a single LISTEN instead of 1 database connection per client. The downside is that now the app code is responsible for tenant isolation instead of the database.
We have very different things in our queues :) An MQ, to me, is a place to put stuff that needs reliable, at-least-once delivery. Like, say, invoices. User never receives that, bad things happen.
There's very little difference between that type of data, and the data in your DB, while it's "active." It's only once the data in the queue is fully ACKed and no longer relevant that the data behaves more ephemerally than first-class DB state.
> and you seem to agree that running it is far simpler OpEx than PG too
I did not say that. In our stack, we run a Redis cluster on top of Kubernetes, because it's nominally semi-stateless. In reality, it's our biggest SPOF. It fights the container daemon around memory allocations (why does a Redis container that uses 2GB memory steady-state need a 64GB memory limit to finish starting up!?) and has weird dynamics around startup readiness (i.e. Redis started from a large AOF-with-RDB-prolog file will start serving requests for just a moment, then immediately stop again to finish loading. This confuses k8s into thinking it's done with startup-probes and is ready for readiness-probes. Which it's not.)
Redis, when configured to be both Highly Available and durable (which is not your use-case, I know) is really just less-robust than a DB, with less management tooling, fewer docs, fewer people who know how to scale it, fewer third-party solutions pre-adapted to expect it, etc. Because, to paraphrase: Redis isn't really in the business of being a database. And yet that's exactly what you end up with, when you try to use Redis as a reliable at-least-once MQ: a database. Just, a kinda sucky one. (One that doesn't even have a concept of users with different privileges; let alone RBAC!)
And yet, that actually is the niche of Redis. It's the durable, client-server cousin to in-process peer-mesh state-sync solutions like Hazelcast/Mnesia. Which makes it kind of weird that it's not particularly tuned at solving the problems in the one niche unique to it.
(I have nothing against Redis! It has non-unique use-cases where it works great, e.g. when it's being used as "memcached with more data structures", or as a coordination/locking server instead of Zookeeper, etc. A robust reliable MQ it is not. My argument is that your average RDBMS is closer to being a robust, reliable MQ than Redis is. Though, of course, you'll get the best robustness+reliability for that use-case out of a "real" MQ — Kafka or NATS or something.)
> but rather that the libraries for DB queues are relying on the non-SQL-standard features/behavior to mock out a full queue functionality
They don't have to, because everything you need to implement reliable queuing semantics is already part of [the lowest-common-denominator part of the] SQL standard [that everything that calls itself an RDBMS implements.]
These libs have DB-specific adapters for two reasons:
1. There's no such wire protocol as "SQL" — every RDBMS speaks its own wire protocol to convey your SQL statements to the database and convey result-sets back; and so, unless you're using a language with a common API abstraction for RDBMS client libs to be built against (e.g. JDBC), then each such client lib ends up presenting its own distinct API to callers, which needs a separate adapter written for it.
2. with SQL, reliable queuing is "easy", but knowing about enqueued items (with as minimal an overhead as possible, scanning as few additional items as possible) is hard. SQL says nothing about how to get push-notification-to-the-client type information out of the DB, and so every DBMS implements its own mechanism for this. These are pure optimizations using pure-ephemeral code, though; the schema of the queue in the DB remains the same either way, such that the data could be exported from one RDBMS to a .sql file, imported from that .sql file to a different RDBMS, and the only thing needed on the client side would be switching adapters.
(In theory, presuming a common wire-protocol abstraction interface lib like JDBC, you could have a "standard SQL" adapter. It'd just be annoyingly low-throughput. Like using HTTP polling when you could be using WebSockets.)
ok, but how do you overbook a flight without booking at least one of the seats twice?
How do you write tests? With what data are you testing?
How do you handle backwards compatibility? (For the code? For the data?)
Do you do this upgrade during deployment? Can you do blue/green deployments? What about canary deployments?
Is this transaction where you are dropping the old and creating the new part of the code or part of the database? (the actual code that does this)
How do you profile the performance of the said code?
SQS without FIFO works about like this AFAIK.
I agree with the gist of everything you are saying though, particularly the point about signaling. As soon as you want to start keeping track of a process that state is going to be stored somewhere outside the queues.
If you ever really need to that's one way to do it but consider I'm imagining 50-100k state updates per second as being reasonable depending on how the data and queries are laid out. You can always use NOTIFY as discussed to cut down on the amount of worker queries for new work before having to completely push signaling to a separate system(which would likely still require occasional polling and/or a background process to re-signal work that hasn't been picked up).
It's interesting to consider that AWS Step Functions charges(or did charge) by the state transition.
Problem was there wasn't a good answer to "How much redis does your team need to know to put it in production".
We thought we knew it well enough, we thought we knew what we were getting into, and we thought so many others are using it for this, we should be good. That is makes a difference, clearly.
In most other languages, you’re sending everything to the queue. With Elixir you only need a small subset of background work to go to a queue, which is usually work that would stress the database.
How is relational storage engine with support for transactions, document/json, ACID, hot-standby "the wrong data model" for a queue?
Read here https://status.circleci.com/incidents/8rklh3qqckp1
Early stage startups die because of lack of PMF. Diverting focus and resources away from finding PMF kills companies. Most companies should focus on the product, tech debt be damned.
Right, it doesn't have to be. Blockchain tech such as Bitcoin is effectively this sort of thing. The idealized type of queue worker would be a pure, referential transparent function that crunches some data and returns an output. More difficult to write but still acceptable: idempotent workers.
> the worker is assumed to be stalled and a new child request is created
This may work for your use case. I'd be worried about an infinite process loop here. Where a problematic job continually stalls a child process and gets entered back into the queue forever, because there needs to be logic to raise an exception and say: "hey, take a look at this job... something may be wrong with it" and then put that job in a holding area. This is also why I'm a big fan of Erlang, which has all of this batteries included with supervisors (they really were 20+ years ahead of their time on this stuff).
> I think there's huge difference in the needs of different implementations of queues
Yes, and people often aren't sure what their needs are. I've worked with RabbitMQ on a system that almost certainly didn't need it. And I've worked on systems that were nothing more than a file that lands via FTP on a folder and is moved (atomically, via /bin/mv) to indicate the job is pending. Using a file system as a queue is another interesting take on this sort of thing. No database even needed.
But you only need one connection for LISTEN per database, total. So I'm confused why this is made out to be a big issue.
I've been very happy with it.
You can use pgtap to write automated unit tests. For profiling, there are explain, autoexplain, and plprofiler.
Blue/Green and Canary deployments are not currently possible with Postgres. On the other hand, Postres has transactional DDL, which means there is no downtime during code deployment and automatic rollback if something goes wrong.
The database is only for data and for stored procedures or functions, which are stored in separate schemas. Your deployment scripts, migration scripts, test scripts and everything else is not stored in the database, but in your source control system, e.g. Git.
For everything else, just use conventional software engineering practices. There is no reason to treat SQL code differently than Ruby or Java code.
You are absolutely right about that. However, in this case we are talking about background workers. I would argue that background workers, such as an image resizing worker, are typically CPU-bound and do not perform high-volume OLTP queries. Therefore, a background worker does not require multiple database connections per thread as a web server would.
- storing relational data (duh)
- storing JSON documents with Postgres' JSONB support - it really is very good, and being able to query relational and document data in the same query is wonderful
- storing key/value type data, where I only need to store a small amount of such data - seems silly to spin up Redis for such a small requirement
- time-series data - TimescaleDB is amazing. Performance may not be on par with a highly tuned schema with a purpose-built time series database, but it's still very, very good. It's fast even with billions of rows, has data compression, and it's really nice to be able to be able to query it just like any other Postgres tables. And the TimescaleDB folks are really helpful on Slack and GitHub. I'm a huge fan of TimescaleDB, and think it's more than adequate for a lot of time-series use cases
- full text search - Postgres shines here too! It's not as powerful as the likes of Elasticsearch, but it's still very good, and very fast. And Elasticsearch is not trivial or cheap to setup and maintain
For queues and pub/sub, RabbitMQ is my go-to solution.
It can be tricky to configure it you want clustering, or have fancy requirements (e.g. client certificate authentication), but once running, it's solid.
I mitigate this in several ways, first, only one worker can work on a job or child job. At worst, one worker is stuck in a loop. There's also a maximum limit of retries. So a job is never infinitely retried. Increasing the amount of stall time per request would be trivial but it's never become an issue. Monitoring potential infinite jobs is as simple as querying for jobs that reach the limit.
I'm sure Erlang or other actor model type implementations would handle this incredibly well at scale but it seems to me that just doing in the database works good enough for at least certain workloads.
As always, it comes down to a question of the right tool for the job. The advantage of the database is that it's mostly likely already in the stack. I don't doubt that there are situations where it's the wrong tool. That's why it's interesting to know at which point it's not a good solution, so that one can make a good decision.
seriously, you have the actual data, you have the store procedures and you have the code. Each has their own version. I have never seen this work in a production environment. It's possible that things evolved and there is better tooling since I last tried performing this stunt.
If it works for you that's great and maybe you should write some sort of blog post (do people still blog?) describing the setup and allowing others to either 1) replicate and use it 2) poke holes in it.
How do you handle this in other languages?
How does this nonsense add anything to the discussion?
Thus neither are running in the same transaction as anything else. The SELECT will never block, since it is not specified as "FOR update". The update itself specified the key.
On the other hand, the SKIP LOCKED approach is really nice for allowing the queue processing to be one single transaction that commits atomically, and thus nicer semantics than a timeout based retry strategy. I do like that approach. The only downside is that not all RDBMS support those semantics, but for an app that only needs to support Postgres it does look like the smart approach.
The original article handles the first argument: Postgres doesn't need polling. Instead it provides a notify mechanism, that informs the application when the table changed (something was added or removed from the queue) via the SQL NOTIFY statement.
For the second point it also provides a solution: since optimization for reading is not needed anymore with the NOTIFY statement, the trade-off like different: we now need an efficient way to write. For this the article provides an efficient update statement with special lock behavior. This helps to make writes efficient, too.
It looks like both points from the blog post you linked are handled in the original article.
> I struggle to consider using it as a data store these days as well with other options on the table.
It sounds like you wouldn't use it for any purposes.
Sometimes it doesn't work out and that's when airlines bleed. I was flying home one day from some US city (was it Denver? the years and cities blend together) when a Drupal conference and some major sports thing ended the same day and the airport was a total gong show. United first offered a paltry 120 USD but when they raised to 400 USD to go home the next day instead my hand went up -- and I was such a greenhorn I didn't know they will pay for the hotel, too. A buddy pocketed no less than 1200 USD from Delta.