Skip to main content
Table of contents

Backups

Last updated: 10 Aug 2020

Backup and restore Elasticsearch indices

GOV.UK uses AWS Managed Elasticsearch which takes daily snapshots of the cluster as part of the managed service. These are stored in a S3 bucket that is not made available to us. Restoration is done by making HTTP requests to the _snapshot endpoint.

To restore a snapshot, follow these steps:

  1. SSH to a search box:

     gds govuk connect ssh -e integration search
    
  2. Query the _snapshot endpoint of Elasticsearch to get the snapshot repository name:

     govuk_setenv search-api \
     bash -c 'curl "$ELASTICSEARCH_URI/_snapshot?pretty"'
    
  3. Query the _all endpoint to identify the available snapshots in the named repository:

     govuk_setenv search-api \
     bash -c 'curl "$ELASTICSEARCH_URI/_snapshot/<repository-name>/_all?pretty"'
    
  4. If an index already exists with the same name as the one being restored, delete the existing index:

     govuk_setenv search-api \
     bash -c 'curl -XDELETE "$ELASTICSEARCH_URI/<index-name>"'
    
  5. Restore the index from the snapshot:

     govuk_setenv search-api \
     bash -c 'curl -XPOST -H 'Content-Type: application/json' "$ELASTICSEARCH_URI/_snapshot/<repository-name>/<snapshot-id>/_restore"  -d "{\"indices\": \"<index-name>\"}"'
    

Further information about Elasticsearch snapshots can be found in the AWS documentation

After a restore has taken place, you will need to fix the out-of-date search indices following the restore, since any changes made in publishing apps since the backup was taken will be missing.