[BUG] Policy deletes indices too early

[torbenaa]

Describe the bug

State management policy deletes indices too early, ignoring the 120 days (120d) specified on the policy step.

Related component

Other

To Reproduce

Create the index policy :

PUT _plugins/_ism/policies/production-log-policy
{
  "policy" : {
    "policy_id" : "production-log-policy",
    "description" : "Delete after 4 month.",
    "default_state" : "hot",
    "states" : [
      {
        "name" : "hot",
        "actions" : [
          {
            "retry" : {
              "count" : 3,
              "backoff" : "exponential",
              "delay" : "1m"
            },
            "index_priority" : {
              "priority" : 100
            }
          }
        ],
        "transitions" : [
          {
            "state_name" : "warm",
            "conditions" : {
              "min_index_age" : "14d"
            }
          }
        ]
      },
      {
        "name" : "warm",
        "actions" : [
          {
            "retry" : {
              "count" : 3,
              "backoff" : "exponential",
              "delay" : "1m"
            },
            "index_priority" : {
              "priority" : 50
            }
          }
        ],
        "transitions" : [
          {
            "state_name" : "delete",
            "conditions" : {
              "min_index_age" : "120d"
            }
          }
        ]
      },
      {
        "name" : "delete",
        "actions" : [
          {
            "retry" : {
              "count" : 3,
              "backoff" : "exponential",
              "delay" : "1m"
            },
            "delete" : { }
          }
        ],
        "transitions" : [ ]
      }
    ],
    "ism_template" : [
      {
        "index_patterns" : [
          "logstash-xxx-prod*"
        ],
        "priority" : 1
      }
    ]
  }
}

Expected behavior

I would expect the indices to go into "warm" state after 14 days, and then after 120d be deleted.
The policy deletes all matching indices after 14 days !

Additional Details

Running Opensearch 3-node cluster on Kubernetes based on the ofiicial helm charts.

Helm chart version is opensearch-2.17.2
Opensearch version is 2.11

[dblock]

Looks like a bug.

Catch All Triage - 1 2 3 4 5