Backing Up Elasticsearch

Elasticsearch replicas protect against a node going down, but they won’t help you with a catastrophic failure. Only good backup practices can help you then.

Back up and restore your Elasticsearch cluster in three steps:

  1. Configure a repository

  2. Make a snapshot of the cluster

  3. Restore from the snapshot

For more detailed information, refer to the Elasticsearch administration guide, and in particular to the documentation on the Snapshot and Restore module.

Creating a Repository

First create a repository to store your snapshots. Several repository types are supported:

  • Shared file system, such as a Network File System or NAS
  • Amazon S3
  • HDFS (Hadoop Distributed File System)
  • Azure Cloud

If using a shared file system repository type, first register the path to the shared file system in each node’s elasticsearch.yml using the path.repo setting.

path.repo: ["path/to/shared/file/system/"]

Once the path to the folder hosting the repository is registered (make sure the folder exists), create the repository with a PUT command. For example,

curl -X PUT "localhost:9200/_snapshot/test_backup" -H 'Content-Type: application/json' -d'
  "type": "fs",
  "settings": {
    "location": "/path/to/shared/file/system/"

Replace localhost:9200 with the proper hostname:port combination for your system, replace test_backup with the name of the repository to create, and use the absolute path to your shared file system in the location.

If the repository is set up successfully, you see this message:


Once the repository exists, you can start creating snapshots.

Taking Snapshots of the Cluster

The easiest snapshot approach is to create a snapshot of all the indexes in your cluster. For example,

curl -XPUT localhost:9200/_snapshot/test_backup/snapshot_1

If {"accepted":true} appears in the terminal, the snapshot was a success.

It’s possible to be more selective when taking snapshots. For example, if you use LES Monitoring, you can exclude the monitoring indexes. Explicitly declare the indexes to include in the snapshot. For example,

curl -XPUT localhost:9200/_snapshot/test_backup/snapshot_2
{ "indices": "liferay-0,liferay-20116" }

Note: For a list of all the Elasticsearch indexes, use this command:

curl -X GET "localhost:9200/_cat/indices?v"

This shows the index metrics:

health status index         uuid                   pri rep docs.count docs.deleted store.size
green  open   liferay-20099 obqiNE1_SDqfuz7rincrGQ   1   0        195            0    303.1kb        303.1kb
green  open   liferay-47206 3YEjtye1S9OVT0i0EZcXcw   1   0          7            0     69.7kb         69.7kb
green  open   liferay-0     shBWwpkXRxuAmGEaE475ug   1   0        147            1    390.9kb        390.9kb

It’s important to note that Elasticsearch uses a smart snapshots approach. To understand what that means, consider a single index. The first snapshot includes a copy of the entire index, while subsequent snapshots only include the delta between the first, complete index snapshot and the current state of the index.

Eventually you’ll end up with a lot of snapshots in your repository, and no matter how cleverly you name the snapshots, you may forget what some snapshots contain. For this purpose, the Elasticsearch API provides getting information about any snapshot. For example:

curl -XGET localhost:9200/_snapshot/test_backup/snapshot_1



There’s lots of useful information here, including which indexes were included in the snapshot.

If you want to get rid of a snapshot, use the DELETE command.

curl -XDELETE localhost:9200/_snapshot/test_backup/snapshot_1

You might trigger creation of a snapshot and regret it (for example, you didn’t want to include all the indexes in the snapshot). If your snapshots contain a lot of data, this can cost time and resources. To cancel the ongoing creation of a snapshot, use the same DELETE command. The snapshot process is terminated and the partial snapshot is deleted from the repository.

Restoring from a Snapshot

What good is a snapshot if you can’t use it to restore your search indexes in case of catastrophic failure? Use the _restore API to restore all the snapshot’s indexes:

curl -XPOST localhost:9200/_snapshot/test_backup/snapshot_1/_restore

Restore only specific indexes from a snapshot by passing in the indices option, and rename the indexes using the rename_pattern and rename_replacement options:

curl -XPOST
    "indices": "liferay-20116",
    "rename_pattern": "liferayindex_(.+)",
    "rename_replacement": "restored_liferayindex_$1"

This restores only the index named liferay-20116index_1 from the snapshot. The rename... settings specify that the beginning liferayindex_ are replaced with restored_liferayindex_, so liferay-20116index_1 becomes restored_liferay-20116index_1.

As with the process for taking snapshots, an errant restored index can be canceled with the DELETE command:

curl -XDELETE localhost:9200/restored_liferay-20116index_3

Nobody likes catastrophic failure on a production system, but Elasticsearch’s API for taking snapshots and restoring indexes can help you rest easy knowing that your search cluster can be restored if disaster strikes. For more details and options, read Elastic’s documentation on the Snapshot and Restore Module.

« Installing Liferay Enterprise Search SecurityUpgrading to Elasticsearch 6.5 »
Was this article helpful?
0 out of 0 found this helpful