Comments on: Minimizing Downtime from Lengthy AWS Outages https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/ Tue, 06 Aug 2019 20:24:18 +0000 hourly 1 https://wordpress.org/?v=6.5.2 By: Anil https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1835972 Wed, 12 Jun 2013 17:28:35 +0000 https://www.percona.com/blog/?p=11381#comment-1835972 Please post your experience on Percona XtraDB Cluster (PXC) on AWS. I would help us since we want to explore this option.

]]> By: Aldo https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1281028 Tue, 05 Mar 2013 12:23:24 +0000 https://www.percona.com/blog/?p=11381#comment-1281028 Thanks for the post! the benefits of using AWS are enough reasons to have a good contingency plan I’ll suggest this topic to be included in the course on AWS.

]]> By: William https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1079186 Thu, 25 Oct 2012 14:48:39 +0000 https://www.percona.com/blog/?p=11381#comment-1079186 @Jeremiah Avoiding the AWS will help you avoid AWS outages, but then you’ll have your own infrastructure problems. Mixed in with a few AWS specifics, there are some geographic distribution tips as well.

]]> By: Jeremiah https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1079061 Thu, 25 Oct 2012 11:45:54 +0000 https://www.percona.com/blog/?p=11381#comment-1079061 Not using “The Cloud” seems to be the obvious way to avoid the impact of outages like this.

]]> By: Jacky https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1078618 Thu, 25 Oct 2012 03:18:23 +0000 https://www.percona.com/blog/?p=11381#comment-1078618 Just curious, have anyone done performance test for PXC in single availability zone, multi availability zone, and multi region? This will be a good insight on how well PXC do in HA setup of EC2

]]> By: Justin Swanhart https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1078512 Thu, 25 Oct 2012 01:36:38 +0000 https://www.percona.com/blog/?p=11381#comment-1078512 I like the idea of using two PXC clusters in different regions. Use at least three nodes in each region and place each node into a different AZ. The extra latency on commit won’t be too bad as the latency between AZs is fairly low.

You can use normal asynch replication between regions. I suggest this for two reasons. First, you don’t want an Internet transit problem (suddenly very slow connections, or lots of packet loss etc) to slow down writing into your primary region.

Second, while PXC performs much better when geographically distributed compared to semi-sync replication, the extra latency on commit may not be acceptable for your application if you extend your cluster across regions, particularly if you want your primary region to be in the US and the secondary region in Europe.

Some other things to keep in mind:
Make sure you encrypt your traffic between zones and regions. Use SSL, stunnel, openvpn, etc between your nodes.

Make sure you run backups in each of the regions.

Make sure you test failures and failure modes in each region.

]]> By: Karl https://www.percona.com/blog/minimizing-downtime-from-lengthy-aws-outages/#comment-1078477 Thu, 25 Oct 2012 00:47:05 +0000 https://www.percona.com/blog/?p=11381#comment-1078477 And avoid North Virginia at all cost. Most of the issues are coming from this region, since it’s their first one.

]]>