Elasticsearch 1.7 to 2.4 upgrade and slow indexing

We recently upgraded elasticsearch 1.7 to 2.4 . Our batch indexing used to take 20 ms before upgrade. After upgrade it’s taking 500-800 ms. Please let me know if there is known issue with es 2.4 or if there is anything that can be performed to speed up the indexing. Our batch size is 1000.

There was a change in Elasticsearch 2.0 to guarantee durability for each document written, that was not there in 1.x - that might be a contributing factor.

Apart from that you have not specified any more info, which would be useful to take a look. Versions (ES, JVM), what exactly you are doing. Is it the same load? Is it happening constantly? What about CPU/IO load? Are you using the exact same machines, etc... please take your time and write up a proper report describing your environment, otherwise it is very hard to ehlp.

--Alex

  1. ES Version 2.4.4
  2. Java version 1.7.0_80
  3. Load havn't changed.
  4. We did side by side upgrade (new machines with the same configuration as old ones).
  5. It is happening constantly.
  6. CPU utilization is less than 15% and we get 25k IOPS and it peaks at 50k.

@spinscale I updated the required information.

have you used async replication, which has been removed in 2.0?

@spinscale We havn't used async replication. We resolved this issue by migrating the indices to different AWS instance type. Thanks for your help.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.