Getting started with snapshot lifecycle management

Let’s get started with snapshot lifecycle management (SLM) by working through a hands-on scenario. The goal of this example is to automatically back up Elasticsearch indices using the snapshots every day at a particular time.

Before starting, it’s important to understand the privileges that are needed when configuring SLM if you are using the security plugin. There are two built-in cluster privileges that can be used to assist: manage_slm and read_slm. It’s also good to note that the create_snapshot permission allows taking snapshots even for indices the role may not have access to.

An example of configuring an administrator role for SLM follows:

POST /_security/role/slm-admin
{
  "cluster": ["manage_slm", "create_snapshot"],
  "indices": [
    {
      "names": [".slm-history-*"],
      "privileges": ["all"]
    }
  ]
}

Or, for a read-only role that can retrieve policies (but not update, execute, or delete them), as well as only view the history index:

POST /_security/role/slm-read-only
{
  "cluster": ["read_slm"],
  "indices": [
    {
      "names": [".slm-history-*"],
      "privileges": ["read"]
    }
  ]
}

Before we can set up an SLM policy, we’ll need to set up a snapshot repository where the snapshots will be stored. Repositories can use many different backends, including cloud storage providers. You’ll probably want to use one of these in production, but for this example we’ll use a shared file system repository:

PUT /_snapshot/my_repository
{
  "type": "fs",
  "settings": {
    "location": "my_backup_location"
  }
}

Now that we have a repository in place, we can create a policy to automatically take snapshots. Policies are written in JSON and will define when to take snapshots, what the snapshots should be named, and which indices should be included, among other things. We’ll use the Put Policy API to create the policy.

PUT /_slm/policy/nightly-snapshots
{
  "schedule": "0 30 1 * * ?", 
  "name": "<nightly-snap-{now/d}>", 
  "repository": "my_repository", 
  "config": { 
    "indices": ["*"] 
  }
}

This policy will take a snapshot of every index each day at 1:30AM UTC. Snapshots are incremental, allowing frequent snapshots to be stored efficiently, so don’t be afraid to configure a policy to take frequent snapshots.

In addition to specifying the indices that should be included in the snapshot, the config field can be used to customize other aspects of the snapshot. You can use any option allowed in a regular snapshot request, so you can specify, for example, whether the snapshot should fail in special cases, such as if one of the specified indices cannot be found.

While snapshots taken by SLM policies can be viewed through the standard snapshot API, SLM also keeps track of policy successes and failures in ways that are a bit easier to use to make sure the policy is working. Once a policy has executed at least once, when you view the policy using the Get Policy API, some metadata will be returned indicating whether the snapshot was sucessfully initiated or not.

Instead of waiting for our policy to run, let’s tell SLM to take a snapshot as using the configuration from our policy right now instead of waiting for 1:30AM.

PUT /_slm/policy/nightly-snapshots/_execute

This request will kick off a snapshot for our policy right now, regardless of the schedule in the policy. This is useful for taking snapshots before making a configuration change, upgrading, or for our purposes, making sure our policy is going to work successfully. The policy will continue to run on its configured schedule after this execution of the policy.

GET /_slm/policy/nightly-snapshots?human

This request will return a response that includes the policy, as well as information about the last time the policy succeeded and failed, as well as the next time the policy will be executed.

{
  "nightly-snapshots" : {
    "version": 1,
    "modified_date": "2019-04-23T01:30:00.000Z",
    "modified_date_millis": 1556048137314,
    "policy" : {
      "schedule": "0 30 1 * * ?",
      "name": "<nightly-snap-{now/d}>",
      "repository": "my_repository",
      "config": {
        "indices": ["*"],
      }
    },
    "last_success": { 
      "snapshot_name": "nightly-snap-2019.04.24-tmtnyjtrsxkhbrrdcgg18a", 
      "time_string": "2019-04-24T16:43:49.316Z",
      "time": 1556124229316
    } ,
    "last_failure": { 
      "snapshot_name": "nightly-snap-2019.04.02-lohisb5ith2n8hxacaq3mw",
      "time_string": "2019-04-02T01:30:00.000Z",
      "time": 1556042030000,
      "details": "{\"type\":\"index_not_found_exception\",\"reason\":\"no such index [important]\",\"resource.type\":\"index_or_alias\",\"resource.id\":\"important\",\"index_uuid\":\"_na_\",\"index\":\"important\",\"stack_trace\":\"[important] IndexNotFoundException[no such index [important]]\\n\\tat org.elasticsearch.cluster.metadata.IndexNameExpressionResolver$WildcardExpressionResolver.indexNotFoundException(IndexNameExpressionResolver.java:762)\\n\\tat org.elasticsearch.cluster.metadata.IndexNameExpressionResolver$WildcardExpressionResolver.innerResolve(IndexNameExpressionResolver.java:714)\\n\\tat org.elasticsearch.cluster.metadata.IndexNameExpressionResolver$WildcardExpressionResolver.resolve(IndexNameExpressionResolver.java:670)\\n\\tat org.elasticsearch.cluster.metadata.IndexNameExpressionResolver.concreteIndices(IndexNameExpressionResolver.java:163)\\n\\tat org.elasticsearch.cluster.metadata.IndexNameExpressionResolver.concreteIndexNames(IndexNameExpressionResolver.java:142)\\n\\tat org.elasticsearch.cluster.metadata.IndexNameExpressionResolver.concreteIndexNames(IndexNameExpressionResolver.java:102)\\n\\tat org.elasticsearch.snapshots.SnapshotsService$1.execute(SnapshotsService.java:280)\\n\\tat org.elasticsearch.cluster.ClusterStateUpdateTask.execute(ClusterStateUpdateTask.java:47)\\n\\tat org.elasticsearch.cluster.service.MasterService.executeTasks(MasterService.java:687)\\n\\tat org.elasticsearch.cluster.service.MasterService.calculateTaskOutputs(MasterService.java:310)\\n\\tat org.elasticsearch.cluster.service.MasterService.runTasks(MasterService.java:210)\\n\\tat org.elasticsearch.cluster.service.MasterService$Batcher.run(MasterService.java:142)\\n\\tat org.elasticsearch.cluster.service.TaskBatcher.runIfNotProcessed(TaskBatcher.java:150)\\n\\tat org.elasticsearch.cluster.service.TaskBatcher$BatchedTask.run(TaskBatcher.java:188)\\n\\tat org.elasticsearch.common.util.concurrent.ThreadContext$ContextPreservingRunnable.run(ThreadContext.java:688)\\n\\tat org.elasticsearch.common.util.concurrent.PrioritizedEsThreadPoolExecutor$TieBreakingPrioritizedRunnable.runAndClean(PrioritizedEsThreadPoolExecutor.java:252)\\n\\tat org.elasticsearch.common.util.concurrent.PrioritizedEsThreadPoolExecutor$TieBreakingPrioritizedRunnable.run(PrioritizedEsThreadPoolExecutor.java:215)\\n\\tat java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)\\n\\tat java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)\\n\\tat java.base/java.lang.Thread.run(Thread.java:834)\\n\"}"
    } ,
    "next_execution": "2019-04-24T01:30:00.000Z", 
    "next_execution_millis": 1556048160000
  }
}

If you’re following along, the returned SLM policy shouldn’t have a last_failure field - it’s included above only as an example. You should, however, see a last_success field and a snapshot name. If you do, you’ve successfully taken your first snapshot using SLM!

While only the most recent sucess and failure are available through the Get Policy API, all policy executions are recorded to a history index, which may be queried by searching the index pattern .slm-history*.

That’s it! We have our first SLM policy set up to periodically take snapshots so that our backups are always up to date. You can read more details in the SLM API documentation and the general snapshot documentation.