How to (offsite) backup and recover snapshots

Hi - have a ES cluster consisting of lets say 20 nodes for arguments sake. Of course the indexes and shards are spread across the nodes... now lets say I do a snapshot.
It looks like each node snapshots the data on itself to the local repository on that node.

Snapshotting and restoring seems to work just fine ... .but what if I need to make offsite backups of the snapshots ? Do I really need to onto all nodes and copy its backup repo away ?
Which also has me wondering - as I havent tested this - How would one then go about restoring a snapshot if the cluster does not have exactly the same node configuration.

I realise the ideal situation is that your data is replicated in such a way that you would not need to "recover" a cluster - the reality however is that I need to cater for a "disaster recovery" type situation where I might lose a large part (or the whole cluster) cluster in one DC and be able to restore the data in the DR site.

Thanks in advance

No. All snapshots need to be done through the snapshot API, which is the only supported snapshot mechanism, and this requires shared storage that is accessible to all nodes, e.g. NFS, S3, HDFS etc.

You do not need the same size cluster as indices are snapshotted cluster wide. If your clusters have different node types you may need to override this when you restore.

Ah ok .. so I should mount the Same backup repository to All nodes. And they'll all write into the same directory and be ok with that ?
I thought I tried it like that - but ran into issues. There are some nodes in the cluster that have the same name .?

Just to add that there's a process for taking a backup of your snapshot repository described in the manual.

2 Likes

Which version are you using?

ES 7.16 ... although my tests might have been on a slightly older version of 7.x.

But thanx - I guess its an easy enough test for me to do. :+1:

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.