Impact of Amazon EBS outage
Hey guys,
I was wondering if Appharbor was impacted by the Amazon 04/21 - 04/25 outage. If so, any lessons learned or recommendations?
Thanks
Jaspreet
PS: I am making this post private but feel free to make it public if you wish.
Discussions are closed to public comments.
If you need help with AppHarbor please
start a new discussion.
Keyboard shortcuts
Generic
| ? | Show this help |
|---|---|
| ESC | Blurs the current field |
Comment Form
| r | Focus the comment reply box |
|---|---|
| ^ + ↩ | Submit the comment |
You can use Command ⌘ instead of Control ^ on Mac
Support Staff 1 Posted by friism on 26 Apr, 2011 01:29 AM
AppHarbor was not affected by the outage. The outage happened in the US-east region where AppHarbor is also located. The US-east region has multiple availability zones (four at the time of writing). The EBS outage was mostly contained to a single availability zone, although there was initially some contagion to other zones. AppHarbor had no instances running in the affected zone and none of our instances were affected by the contagion.
Had AppHarbor been in the affected region, and had our servers stopped responding, we would have been alerted immediately. Application hosting on affected web servers could have been restored fairly quickly by redeploying applications to backup web servers in an unaffected zone. MySQL databases (including the AppHarbor database) most likely would have been OK, because we run a redundant multi-availability-zone setup. SQL Server database would have taken somewhat longer to get back online.
We have confidence in the Amazon Web Services platform. We are also constantly working on improving redundancy and finetuning recovery procedures to make AppHarbor tolerant to failures in the underlying infrastructure. We are committed to providing a highly available deployment and hosting service.
When we open up for the possibility of deploying your application to multiple instances we would recommend doing so for increased redundancy. Should one of the web servers crash, the load balancer will make sure to fail over to the running instances. We can also recommend using the MySQL database option at this point because it's the only fully redundant shared database option we currently offer.
friism closed this discussion on 26 Apr, 2011 01:30 AM.