Index rollover latency

Question

I have created an ILM policy as follows,

{
  "policy": {
    "phases": {
      "hot": {
        "min_age": "0ms",
        "actions": {
          "rollover": {
            "max_docs": 30
          },
          "set_priority": {
            "priority": 100
          }
        }
      },
      "delete": {
        "min_age": "15m",
        "actions": {
          "delete": {}
        }
      }
    }
  }
}

Index setting as follows,

{
  "index": {
    "lifecycle": {
      "name": "index-roll+doc-over",
      "rollover_alias": "router01"
    },
    "number_of_shards": "1",
    "number_of_replicas": "0"
  }
}

As you have already guessed, I am testing this ILM rollover policy which is why I have set the document size only 30. I am planning to limit the doc size of 1.5 Billion. But I am observing huge latency while creating a rollover index and switching to the new index. It is writing on the current index and not rolling over. For the test I have ingested only 500 lines of log.

Index details

health status index                    uuid                   pri rep docs.count docs.deleted store.size pri.store.size

green  open   router01-000001          654G2ArFQiypDn9U69jedw   1   0        500            0    151.1kb        151.1kb

And at the time of writing this question, the other index (router01-000001) is not created yet.

The the _ilm/explain is following,

{
  "indices" : {
    "router01-000001" : {
      "index" : "router01-000001",
      "managed" : true,
      "policy" : "index-roll+doc-over",
      "lifecycle_date_millis" : 1583819391191,
      "age" : "14.08m",
      "phase" : "hot",
      "phase_time_millis" : 1583819391396,
      "action" : "rollover",
      "action_time_millis" : 1583819706013,
      "step" : "check-rollover-ready",
      "step_time_millis" : 1583819706013,
      "phase_execution" : {
        "policy" : "index-roll+doc-over",
        "phase_definition" : {
          "min_age" : "0ms",
          "actions" : {
            "rollover" : {
              "max_docs" : 30
            },
            "set_priority" : {
              "priority" : 100
            }
          }
        },
        "version" : 1,
        "modified_date_in_millis" : 1583756555777
      }
    }
  }
}

Elasticsearch currently has 2GB heap and logstash has 1GB. Total Memory 4GB.

My question is,

Is the latency expected (May be I misunderstood the rollover concept)
Did I mistake anywhere?
If not than what can I do to improve the latency as there will be 20 millions log will be ingested per day

Index rollover latency

0 Answers0