1

I'm using a AWS ElasticSearch that run fine on a v1.5 cluster and i copy the entire config (shards,nodes,instance types, mapping,data) to new v2.3 cluster.

When i point my DNS to the new v2.3 cluster the CPU usage create spikes that slow down my entire application. When im running in v1.5 cluster, the CPU usage it's almost linear, about 20% of CPU usage.

Obs: Clusters configuration is the same, triple checked!

Amazon support tell us that is a v2.3 bug, and the bugfix will be released as soon as possible...

But i don't found any thing about that bug in release notes!

https://www.elastic.co/guide/en/elasticsearch/reference/2.3/release-notes-2.3.3.html

Any tip?

JoeLoco
  • 2,116
  • 4
  • 31
  • 59
  • 1
    Check out you are using Latest JVM: 1.7u55+. Look for Hot Threads, that tells you which one is causing problem. – rajnish Oct 18 '16 at 13:34
  • of course there will be no information about bugs in release notes :) could you show your queries, that cause the issue? – Mysterion Oct 18 '16 at 13:51

1 Answers1

0

I think that bug is only in AWS Elasticsearch. Read this message from AWS Support:

Hello, Thank you for updating the case. After further investigation, looks like the ES 2.3 version has a bug which causes a CPU spike. An edge case was identified resulting in steady CPU spike. This has been fixed and the service team is rolling out patch. The issue should be resolved soon. I do not have a specific ETA as the release depends on a different team. Best regards, Amazon Web Services We value your feedback. Please rate my response using the link below.

Bruno Rocha da Silva
  • 16
  • 3