Create a snapshot

edit

This guide shows you how to take snapshots of a running cluster. You can later restore a snapshot to recover or transfer its data.

In this guide, you’ll learn how to:

  • Automate snapshot creation and retention with snapshot lifecycle management (SLM)
  • Manually take a snapshot
  • Monitor a snapshot’s progress
  • Delete or cancel a snapshot
  • Back up cluster configuration files

The guide also provides tips for creating dedicated cluster state snapshots and taking snapshots at different time intervals.

Prerequisites

edit
  • To use Kibana’s Snapshot and Restore feature, you must have the following permissions:

  • You can only take a snapshot from a running cluster with an elected master node.
  • A snapshot repository must be registered and available to the cluster.
  • The cluster’s global metadata must be readable. To include an index in a snapshot, the index and its metadata must also be readable. Ensure there aren’t any cluster blocks or index blocks that prevent read access.

Considerations

edit
  • Each snapshot must have a unique name within its repository. Attempts to create a snapshot with the same name as an existing snapshot will fail.
  • Snapshots are automatically deduplicated. You can take frequent snapshots with little impact to your storage overhead.
  • Each snapshot is logically independent. You can delete a snapshot without affecting other snapshots.
  • Taking a snapshot can temporarily pause shard allocations. See Snapshots and shard allocation.
  • Taking a snapshot doesn’t block indexing or other requests. However, the snapshot won’t include changes made after the snapshot process starts.
  • You can take multiple snapshots at the same time. The snapshot.max_concurrent_operations cluster setting limits the maximum number of concurrent snapshot operations.
  • If you include a data stream in a snapshot, the snapshot also includes the stream’s backing indices and metadata.

    You can also include only specific backing indices in a snapshot. However, the snapshot won’t include the data stream’s metadata or its other backing indices.

  • A snapshot can include a data stream but exclude specific backing indices. When you restore such a data stream, it will contain only backing indices in the snapshot. If the stream’s original write index is not in the snapshot, the most recent backing index from the snapshot becomes the stream’s write index.

Automate snapshots with SLM

edit

Snapshot lifecycle management (SLM) is the easiest way to regularly back up a cluster. An SLM policy automatically takes snapshots on a preset schedule. The policy can also delete snapshots based on retention rules you define.

Elasticsearch Service deployments automatically include the cloud-snapshot-policy SLM policy. Elasticsearch Service uses this policy to take periodic snapshots of your cluster. For more information, see the Elasticsearch Service snapshot documentation.

SLM security

edit

The following cluster privileges control access to the SLM actions when Elasticsearch security features are enabled:

manage_slm
Allows a user to perform all SLM actions, including creating and updating policies and starting and stopping SLM.
read_slm
Allows a user to perform all read-only SLM actions, such as getting policies and checking the SLM status.
cluster:admin/snapshot/*
Allows a user to take and delete snapshots of any index, whether or not they have access to that index.

You can create and manage roles to assign these privileges through Kibana Management.

To grant the privileges necessary to create and manage SLM policies and snapshots, you can set up a role with the manage_slm and cluster:admin/snapshot/* cluster privileges and full access to the SLM history indices.

For example, the following request creates an slm-admin role:

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

To grant read-only access to SLM policies and the snapshot history, you can set up a role with the read_slm cluster privilege and read access to the snapshot lifecycle management history indices.

For example, the following request creates a slm-read-only role:

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

Create an SLM policy

edit

To manage SLM in Kibana, go to the main menu and click Stack Management > Snapshot and Restore > Policies. To create a policy, click Create policy.

You can also manage SLM using the SLM APIs. To create a policy, use the create SLM policy API.

The following request creates a policy that backs up the cluster state, all data streams, and all indices daily at 1:30 a.m. UTC.

PUT _slm/policy/nightly-snapshots
{
  "schedule": "0 30 1 * * ?",       
  "name": "<nightly-snap-{now/d}>", 
  "repository": "my_repository",    
  "config": {
    "indices": "*",                 
    "include_global_state": true    
  },
  "retention": {                    
    "expire_after": "30d",
    "min_count": 5,
    "max_count": 50
  }
}

When to take snapshots, written in Cron syntax.

Snapshot name. Supports date math. To prevent naming conflicts, the policy also appends a UUID to each snapshot name.

Registered snapshot repository used to store the policy’s snapshots.

Data streams and indices to include in the policy’s snapshots.

If true, the policy’s snapshots include the cluster state. This also includes all feature states by default. To only include specific feature states, see Back up a specific feature state.

Optional retention rules. This configuration keeps snapshots for 30 days, retaining at least 5 and no more than 50 snapshots regardless of age. See SLM retention and Snapshot retention limits.

Manually run an SLM policy

edit

You can manually run an SLM policy to immediately create a snapshot. This is useful for testing a new policy or taking a snapshot before an upgrade. Manually running a policy doesn’t affect its snapshot schedule.

To run a policy in Kibana, go to the Policies page and click the run icon under the Actions column. You can also use the execute SLM policy API.

POST _slm/policy/nightly-snapshots/_execute

The snapshot process runs in the background. To monitor its progress, see Monitor a snapshot.

SLM retention

edit

SLM snapshot retention is a cluster-level task that runs separately from a policy’s snapshot schedule. To control when the SLM retention task runs, configure the slm.retention_schedule cluster setting.

response = client.cluster.put_settings(
  body: {
    persistent: {
      'slm.retention_schedule' => '0 30 1 * * ?'
    }
  }
)
puts response
PUT _cluster/settings
{
  "persistent" : {
    "slm.retention_schedule" : "0 30 1 * * ?"
  }
}

To immediately run the retention task, use the execute SLM retention policy API.

response = client.slm.execute_retention
puts response
POST _slm/_execute_retention

An SLM policy’s retention rules only apply to snapshots created using the policy. Other snapshots don’t count toward the policy’s retention limits.

Snapshot retention limits

edit

We recommend you include retention rules in your SLM policy to delete snapshots you no longer need.

A snapshot repository can safely scale to thousands of snapshots. However, to manage its metadata, a large repository requires more memory on the master node. Retention rules ensure a repository’s metadata doesn’t grow to a size that could destabilize the master node.

Manually create a snapshot

edit

To take a snapshot without an SLM policy, use the create snapshot API. The snapshot name supports date math.

response = client.snapshot.create(
  repository: 'my_repository',
  snapshot: '<my_snapshot_{now/d}>'
)
puts response
# PUT _snapshot/my_repository/<my_snapshot_{now/d}>
PUT _snapshot/my_repository/%3Cmy_snapshot_%7Bnow%2Fd%7D%3E

Depending on its size, a snapshot can take a while to complete. By default, the create snapshot API only initiates the snapshot process, which runs in the background. To block the client until the snapshot finishes, set the wait_for_completion query parameter to true.

response = client.snapshot.create(
  repository: 'my_repository',
  snapshot: 'my_snapshot',
  wait_for_completion: true
)
puts response
PUT _snapshot/my_repository/my_snapshot?wait_for_completion=true

You can also clone an existing snapshot using clone snapshot API.

Monitor a snapshot

edit

To monitor any currently running snapshots, use the get snapshot API with the _current request path parameter.

response = client.snapshot.get(
  repository: 'my_repository',
  snapshot: '_current'
)
puts response
GET _snapshot/my_repository/_current

To get a complete breakdown of each shard participating in any currently running snapshots, use the get snapshot status API.

response = client.snapshot.status
puts response
GET _snapshot/_status

Check SLM history

edit

To get more information about a cluster’s SLM execution history, including stats for each SLM policy, use the get SLM stats API. The API also returns information about the cluster’s snapshot retention task history.

response = client.slm.get_stats
puts response
GET _slm/stats

To get information about a specific SLM policy’s execution history, use the get SLM policy API. The response includes:

  • The next scheduled policy execution.
  • The last time the policy successfully started the snapshot process, if applicable. A successful start doesn’t guarantee the snapshot completed.
  • The last time policy execution failed, if applicable, and the associated error.
response = client.slm.get_lifecycle(
  policy_id: 'nightly-snapshots'
)
puts response
GET _slm/policy/nightly-snapshots

Delete or cancel a snapshot

edit

To delete a snapshot in Kibana, go to the Snapshots page and click the trash icon under the Actions column. You can also use the delete snapshot API.

response = client.snapshot.delete(
  repository: 'my_repository',
  snapshot: 'my_snapshot_2099.05.06'
)
puts response
DELETE _snapshot/my_repository/my_snapshot_2099.05.06

If you delete a snapshot that’s in progress, Elasticsearch cancels it. The snapshot process halts and deletes any files created for the snapshot. Deleting a snapshot doesn’t delete files used by other snapshots.

Back up configuration files

edit

If you run Elasticsearch on your own hardware, we recommend that, in addition to backups, you take regular backups of the files in each node’s $ES_PATH_CONF directory using the file backup software of your choice. Snapshots don’t back up these files. Also note that these files will differ on each node, so each node’s files should be backed up individually.

The elasticsearch.keystore, TLS keys, and SAML, OIDC, and Kerberos realms private key files contain sensitive information. Consider encrypting your backups of these files.

Back up a specific feature state

edit

By default, a snapshot that includes the cluster state also includes all feature states. Similarly, a snapshot that excludes the cluster state excludes all feature states by default.

You can also configure a snapshot to only include specific feature states, regardless of the cluster state.

To get a list of available features, use the get features API.

response = client.features.get_features
puts response
GET _features

The API returns:

{
  "features": [
    {
      "name": "tasks",
      "description": "Manages task results"
    },
    {
      "name": "kibana",
      "description": "Manages Kibana configuration and reports"
    },
    {
      "name": "security",
      "description": "Manages configuration for Security features, such as users and roles"
    },
    ...
  ]
}

To include a specific feature state in a snapshot, specify the feature name in the feature_states array.

For example, the following SLM policy only includes feature states for the Kibana and Elasticsearch security features in its snapshots.

PUT _slm/policy/nightly-snapshots
{
  "schedule": "0 30 2 * * ?",
  "name": "<nightly-snap-{now/d}>",
  "repository": "my_repository",
  "config": {
    "indices": "*",
    "include_global_state": true,
    "feature_states": [
      "kibana",
      "security"
    ]
  },
  "retention": {
    "expire_after": "30d",
    "min_count": 5,
    "max_count": 50
  }
}

Any index or data stream that’s part of the feature state will display in a snapshot’s contents. For example, if you back up the security feature state, the security-* system indices display in the get snapshot API's response under both indices and feature_states.

Dedicated cluster state snapshots

edit

Some feature states contain sensitive data. For example, the security feature state includes system indices that may contain user names and encrypted password hashes. Because passwords are stored using cryptographic hashes, the disclosure of a snapshot would not automatically enable a third party to authenticate as one of your users or use API keys. However, it would disclose confidential information, and if a third party can modify snapshots, they could install a back door.

To better protect this data, consider creating a dedicated repository and SLM policy for snapshots of the cluster state. This lets you strictly limit and audit access to the repository.

For example, the following SLM policy only backs up the cluster state. The policy stores these snapshots in a dedicated repository.

PUT _slm/policy/nightly-cluster-state-snapshots
{
  "schedule": "0 30 2 * * ?",
  "name": "<nightly-cluster-state-snap-{now/d}>",
  "repository": "my_secure_repository",
  "config": {
    "include_global_state": true,                 
    "indices": "-*"                               
  },
  "retention": {
    "expire_after": "30d",
    "min_count": 5,
    "max_count": 50
  }
}

Includes the cluster state. This also includes all feature states by default.

Excludes regular data streams and indices.

If you take dedicated snapshots of the cluster state, you’ll need to exclude the cluster state from your other snapshots. For example:

PUT _slm/policy/nightly-snapshots
{
  "schedule": "0 30 2 * * ?",
  "name": "<nightly-snap-{now/d}>",
  "repository": "my_repository",
  "config": {
    "include_global_state": false,    
    "indices": "*"                    
  },
  "retention": {
    "expire_after": "30d",
    "min_count": 5,
    "max_count": 50
  }
}

Excludes the cluster state. This also excludes all feature states by default.

Includes all regular data streams and indices.

Create snapshots at different time intervals

edit

If you only use a single SLM policy, it can be difficult to take frequent snapshots and retain snapshots with longer time intervals.

For example, a policy that takes snapshots every 30 minutes with a maximum of 100 snapshots will only keep snapshots for approximately two days. While this setup is great for backing up recent changes, it doesn’t let you restore data from a previous week or month.

To fix this, you can create multiple SLM policies with the same snapshot repository that run on different schedules. Since a policy’s retention rules only apply to its snapshots, a policy won’t delete a snapshot created by another policy.

For example, the following SLM policy takes hourly snapshots with a maximum of 24 snapshots. The policy keeps its snapshots for one day.

PUT _slm/policy/hourly-snapshots
{
  "name": "<hourly-snapshot-{now/d}>",
  "schedule": "0 0 * * * ?",
  "repository": "my_repository",
  "config": {
    "indices": "*",
    "include_global_state": true
  },
  "retention": {
    "expire_after": "1d",
    "min_count": 1,
    "max_count": 24
  }
}

The following policy takes nightly snapshots in the same snapshot repository. The policy keeps its snapshots for one month.

PUT _slm/policy/daily-snapshots
{
  "name": "<daily-snapshot-{now/d}>",
  "schedule": "0 45 23 * * ?",          
  "repository": "my_repository",
  "config": {
    "indices": "*",
    "include_global_state": true
  },
  "retention": {
    "expire_after": "30d",
    "min_count": 1,
    "max_count": 31
  }
}

Runs at 11:45 p.m. UTC every day.

The following policy creates monthly snapshots in the same repository. The policy keeps its snapshots for one year.

PUT _slm/policy/monthly-snapshots
{
  "name": "<monthly-snapshot-{now/d}>",
  "schedule": "0 56 23 1 * ?",            
  "repository": "my_repository",
  "config": {
    "indices": "*",
    "include_global_state": true
  },
  "retention": {
    "expire_after": "366d",
    "min_count": 1,
    "max_count": 12
  }
}

Runs on the first of the month at 11:56 p.m. UTC.