Let’s jump into index lifecycle management (ILM) by working through a hands-on scenario. This section will leverage many new concepts unique to ILM that you may not be familiar with. The following sections will explore these in more details.

The goal of this example is to set up a set of indices that will encapsulate the data from a time series data source. We can imagine there is a system like Filebeat that continuously indexes documents into our writing index. We wish to roll over the index after it reaches a size of 50 gigabytes, or has been created 30 days ago, and then delete the index after 90 days.

Setting up a policy

There are many new features introduced by ILM, but we will only focus on a few that are needed for our example. For starters, we will use the Put Policy API to define our first policy. Lifecycle policies are defined in JSON and include specific phases and actions.

PUT _ilm/policy/datastream_policy   
{
  "policy": {                       
    "phases": {
      "hot": {                      
        "actions": {
          "rollover": {             
            "max_size": "50GB",
            "max_age": "30d"
          }
        }
      },
      "delete": {
        "min_age": "90d",           
        "actions": {
          "delete": {}              
        }
      }
    }
  }
}

Here we created the policy called datastream_policy which rolls over the index being written to after it reaches 50 gigabytes, or it is 30 days old. The rollover will occur when either of these conditions is true. The index will be deleted 90 days after it is rolled over.

Applying a policy to our index

There are a few ways to associate a policy to an index. Since we wish specific settings to be applied to the new index created from Rollover, we will set the policy via index templates.

PUT _template/datastream_template
{
  "index_patterns": ["datastream-*"],                 
  "settings": {
    "number_of_shards": 1,
    "number_of_replicas": 1,
    "index.lifecycle.name": "datastream_policy",      
    "index.lifecycle.rollover_alias": "datastream"    
  }
}

The above index template introduces a few new settings specific to ILM. The first being index.lifecycle.name. This setting will configure the "datastream_policy" to the index applying this template. This means that all newly created indices prefixed "datastream-" will be managed by our policy. The other setting used here is index.lifecycle.rollover_alias. This setting is required when using a policy containing the rollover action and specifies which alias to rollover on behalf of this index. The intention here is that the rollover alias is also defined on the index.

To begin, we will want to bootstrap our first index to write to.

PUT datastream-000001
{
  "aliases": {
    "datastream": {
      "is_write_index": true
    }
  }
}

When creating our index, we have to consider a few important configurations that tie our index and our policy together correctly. We need to make sure that our index name matches our index template pattern of "datastream-*", which it does. We are using the Rollover Action in our policy, which requires that our index name ends with a number. In our case, we used 000001. This is important so that Rollover can increment this number when naming the new index created from rolling over.

Our index creation request leverages its template to apply our settings, but we must also configure our rollover alias: "datastream". To do this, we take advantage of write indices. This is a way to define an alias to be used for both reading and writing, with only one index being the index that is being written to at a time. Rollover swaps the write index to be the new index created from rollover, and sets the alias to be read-only for the source index.

Checking progress

Now that we have an index managed by our policy, how do we tell what is going on? Which phase are we in? Is something broken? This section will go over a few APIs and their responses to help us inspect our indices with respect to ILM.

With the help of the Explain API, we can know things like which phase we’re in and when we entered that phase. The API will also provide further info if errors occurred, or if we are blocked on certain checks within actions.

GET datastream-*/_ilm/explain

The above request will retrieve ILM execution information for all our managed indices.

{
  "indices": {
    "datastream-000001": {
      "index": "datastream-000001",
      "managed": true,                           
      "policy": "datastream_policy",             
      "lifecycle_date_millis": 1538475653281,
      "phase": "hot",                            
      "phase_time_millis": 1538475653317,
      "action": "rollover",                      
      "action_time_millis": 1538475653317,
      "step": "attempt-rollover",                
      "step_time_millis": 1538475653317,
      "phase_execution": {
        "policy": "datastream_policy",
        "phase_definition": {                    
          "min_age": "0ms",
          "actions": {
            "rollover": {
              "max_size": "50gb",
              "max_age": "30d"
            }
          }
        },
        "version": 1,                            
        "modified_date_in_millis": 1539609701576
      }
    }
  }
}

You can read about the full details of this response in the explain API docs. For now, let’s focus on how the response details which phase, action, and step we’re in. We are in the "hot" phase, and "rollover" action. Rollover will continue to be called by ILM until its conditions are met and it rolls over the index. Afterwards, the original index will stay in the hot phase until 90 more days pass and it is deleted in the delete phase. As time goes on, new indices will be created and deleted. With datastream-000002 being created when the index mets the rollover conditions and datastream-000003 created after that. We will be able to search across all of our managed indices using the "datastream" alias, and we will be able to write to our to-be-rolled-over write indices using that same alias.

That’s it! We have our first use-case managed by ILM.

To learn more about all our APIs, check out ILM APIs.