Lifecycle policy setting index read-only even with the option disabled

Elastic Stack Elasticsearch
1

Hello,

I have a hot-warm architecture and a couple of ILM that will move the index from the hot nodes to the warm nodes based on the age.

I have an index that is created daily and is moved to the warm nodes after 7 days, to do this I created a Lifecycle Policy using the Kibana UI where I set the number of days and I checked the force merge option, I didn't checked the read only option.

It is possible that for some cases I can get documents that need to be indexed in these indices even after they were moved to the warm nodes, that's why I didn't checked the read only option, but I'm getting some FORBIDDEN/8/index write (api) errors in logstash and when I looked at the index they were set to read only ("index.blocks.write": "true").

I'm using version 7.12.1 currently, an update is planned but no date defined yet.

Is this a bug? What could be setting the index to read only?

This is my policy:

{
  "redacted-ilm-policy" : {
    "version" : 5,
    "modified_date" : "2021-09-19T14:48:23.390Z",
    "policy" : {
      "phases" : {
        "hot" : {
          "min_age" : "0ms",
          "actions" : {
            "set_priority" : {
              "priority" : 100
            }
          }
        },
        "delete" : {
          "min_age" : "61d",
          "actions" : {
            "delete" : {
              "delete_searchable_snapshot" : true
            }
          }
        },
        "warm" : {
          "min_age" : "7d",
          "actions" : {
            "allocate" : {
              "include" : { },
              "exclude" : { },
              "require" : {
                "node_type" : "warm"
              }
            },
            "forcemerge" : {
              "max_num_segments" : 1
            },
            "set_priority" : {
              "priority" : 50
            }
          }
        }
      }
    }
  }
}

The modified date is recent because I saved it again but no changes were made.

These are the settings of an index in the warm phase with the writes blocked.

{
  "index.blocks.read_only_allow_delete": "false",
  "index.write.wait_for_active_shards": "1",
  "index.mapping.ignore_malformed": "true",
  "index.refresh_interval": "30s",
  "index.blocks.write": "true",
  "index.query.default_field": [
    "message"
  ],
  "index.priority": "50",
  "index.number_of_replicas": "0",
  "index.lifecycle.name": "redacted-ilm-policy",
  "index.routing.allocation.require.node_type": "warm"
}

I was able to temporarily solve this issue by manually setting the index.blocks.write to null in every index, but I expect the ILM to works as defined so I would not need to create a crontab just to set this.

2

Well,

I think it is related to these issues in github:

If you have a force merge action in your Lifecycle Policy it will set the index to read-only even if you do not check the read-only action.

1 Like
3

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