Reduce iops during reindexing

Elastic Stack Elasticsearch
1

At the moment my VMs get i/o throttled due to performing many i/o operations during reindexing but is nowhere near it's limit in terms of bandwidth. Is there any way I could reduce the number of i/o operations going to disk? (maybe store some stuff in memory?)

This isn't really an issue and if need be I can just throw more machines and shards into the mix but it would be nice to not hit these i/o limits. For extra info these are 8 vCPU GCP machines using persistent SSDs (with a large enough capacity to not be limited by it).

2

@magnalite

Here are few settings to play with, if you haven't already

  1. set refresh_interval to -1 for the dest index. Once reindexing completes set it back to null.
  2. See if increasing index_buffer_size helps https://www.elastic.co/guide/en/elasticsearch/reference/current/indexing-buffer.html
  3. Set replica of the dest index to 0. Once reindexing complete set replica to whatever you want.

Even if none of these work, please post your findings. It will help others.

1 Like
3

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