3

I have elastic search 7.7.1 with ILM enabled from logstash, i created an ILM policy with hot and delete phase with 2 seconds everyone of them and 48gb for hot phase. So my question is. Why Elasticsearch very slowly moving the indexses from Hot phase to Delete phase? The ilm explain says that index age was about 30minutes before ilm moved the index to delete phase. I see that it stucks in stage "check-rollover-ready" in hot phase or "wait-for-shard-history-leases" in delete phase. So my question is, why the rollover is going so slow? The indexes is empty, so no data or documents there, but he stucks for that long time and just wait for something. How to fix this?

Elastic ILM

SolarCore
  • 133
  • 2
  • 16
  • 1
    Hi @SolarCore, any updates on this? – Tri Nguyen Aug 13 '20 at 05:01
  • 1
    Same here, very annoying – andrzej1_1 Oct 08 '20 at 13:28
  • 1
    I am very sorry, didn't remember to post my research result here. I did a little research now to remember what i was doing a couple of month ago, and not sure if this thing was the problem, but setting indices.lifecycle.poll_interval to needed interval, fixes this issue. You can look at this parameter in [link](https://www.elastic.co/guide/en/elasticsearch/reference/current/ilm-settings.html#_cluster_level_settings) for setting this setting into cluster settings, look at this page with an example [link](https://github.com/elastic/beats/issues/7935#issue-349244928) – SolarCore Oct 08 '20 at 13:53
  • Thank you for your interest with this issue, i really didn't remember to post my answer here :) – SolarCore Oct 08 '20 at 13:54
  • 1
    @SolarCore Same problem and some solution. I fix this after I set `indices.lifecycle.poll_interval=1m`. I guess your es cluster have too many indices to rollover(or other lifecycle action) and everytime poll action only affect part of them. So poll lifecycle more frequency would rollover indices soon. – Libraco Nov 26 '20 at 10:02

0 Answers0