zlacker

but... the s3 buckets are in us-east-1, and postgres in ap-southeast-2. More regions better than one, for maximum impact with minimum effort.

>>ignora+(OP)
Is there at least one case when whole region went down ;)) ?

>>pojzon+Fo
At least once, in 2012? I remember because just two days before we had fully switched over to AWS and hadn’t done multi-region yet. We got our first downtime in 10 years and there was nothing we could do, unlike when we had the servers in a colo. We were at the mercy of Amazon. After that, we moved everything back to real physical servers.

>>ignora+(OP)
The more regions a service is scattered over the greater the odds are that a single region outage somewhere in AWS will take down the whole service.

Consider the extreme case where your service is scattered over every AWS region: here an outage of any AWS region is guaranteed to take down your service.

Compare that to the case where your service is bound to only one region: then the odds of a single region outage taking down your entire service is reduced to 1 out of however many regions AWS has (assuming each region has an equal chance of suffering an outage).

To guard against outages, the failover service has to be scattered over entirely different regions (or, even better, on an entirely different service provider... which is probably a good idea anyway).

>>pmoria+3G
> The more regions a service is scattered over the greater the odds are that a single region outage somewhere in AWS will take down the whole service.

Agree. I think I should have suffixed a /s to my comment above.

> To guard against outages, the failover service has to be scattered over entirely different regions (or, even better, on an entirely different service provider... which is probably a good idea anyway).

Something, something... the greatest trick the devil (bigcloud) ever pulled...

>>pojzon+Fo
Per the spreadsheet here https://awsmaniac.com/aws-outages/ :

There seem to have been multiple "full" outages in 2011-12 in AWS' us-east-1 region, which, granted, is the oldest AWS region and likely has a bunch of legacy stuff. By "full" outages I mean that a few core services fell over but the entire region become inaccessible due to those core failures.

>>samspe+1Q1
Its over 10 years ago tho. Are there any RECENT full region outages ?

Im forseeing a full downtime in Frankfurt this winter tho. Germany is in really bad position when it comes to electricity.