Elasticsearch Cluster Snapshot & Restore

Posted on 01 Nov 2016 by Eric Oestrich

We recently needed to do a cross cluster snapshot and restore for elasticsearch. We were hosted on Elastic Cloud and found out the hard way that the tooling in place there does not work. We did find out that Elasticsearch has a backup repository system in place that works very well. We would have saved ourselves some time if we had just started with this.

Create a user in AWS IAM

First off create a user in AWS IAM that has access to S3. You can attach the full access or limit down to a single bucket as shown in the Elastic Cloud documentation.

{
  "Statement": [
    {
      "Action": [
        "s3:*"
      ],
      "Effect": "Allow",
      "Resource": [
        "arn:aws:s3:::bucket-name",
        "arn:aws:s3:::bucket-name/*"
      ]
    }
  ]
}

Create the Repository

Create the repository in each cluster. This is also taken from that guide. You also need the repository-s3 plugin installed in each cluster for this to work.

sudo bin/elasticsearch-plugin install repository-s3

curl -X PUT localhost:9200/_snapshot/bucket-name -d '{
  "type": "s3",
  "settings": {
    "bucket": "bucket-name",
    "region": "us-east1",
    "access_key": "AKIAYOURKEYHERE",
    "secret_key": "secret-key",
    "compress": true
  }
}'

Snapshot

On the old cluster, create a snapshot. You can check on the status as it processes.

curl -X PUT localhost:9200/_snapshot/bucket-name/snapshot-backup-name
curl -X GET localhost:9200/_snapshot/bucket-name/snapshot-backup-name/_status

There are a lot of options you can provide to the snapshot, including limiting to certain indices.

Restore

On the new cluster, restore the snapshot. You can easily view the status of the restore with regular elasticsearch monitoring tools. The index health with be shown as shards come online.

curl -X POST localhost:9200/_snapshot/bucket-name/snapshot-backup-name/_restore -d '{
  "indices": "one-index"
}'

Conclusion

Hopefully this is of use to others. Snapshotting and restoring manually is a very simple process and was much easier than trying to figure out a custom solution from your elasticsearch host.

comments powered by Disqus
Eric Oestrich
I am:
All posts
Creative Commons License
This site's content is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License unless otherwise specified. Code on this site is licensed under the MIT License unless otherwise specified.