1

I have an installation of Elasticsearch without persistence running and I need to back it up so we can move it to a persistent setup.

What are safe ways to do this and keep the data in the ES database?

It is running and I want to avoid shutting it down.

It is running in a container and I think the entrypoint starts the server. So stopping the cluster my stop my pod and get rid of all my data...

  • 1
  • @Tensibai it looks like you need to define repos to do this and that seems to require restarting the cluster... – David West Aug 30 '18 at 16:59
  • I don't understand why you think that, you do a put request to create a local (in container) repo, another put to create the snapshot and then docker cp this location to host, shut you cluster down, bring it back up with volume mounts or whatever, docker cp to container and restore the snapshot... – Tensibai Aug 30 '18 at 18:04
  • @Tensibai we just moved our cluster. we did not back up the data and just took the loss. now it has persistance. The whole reason we couldn't use snapshots was because that required shutting the cluster down and bringing that back up. With our setup that got rid of the data :) – David West Sep 1 '18 at 17:07
  • You're not willing to accept that a snapshot can be take locally to the container and then copied with docker commands. We can't do anything more for you – Tensibai Sep 1 '18 at 21:50
1

ElasticSearch can backup itself when told to do so via an API call. You just need to configure a destination, and optionally credentials, for example you make ElasticSearch create snapshots of its database in AWS S3.

https://www.elastic.co/guide/en/elasticsearch/reference/current/modules-snapshots.html

You can also automate these calls using a schedule with the Elastic curator tool.

https://www.elastic.co/guide/en/elasticsearch/client/curator/5.5/snapshot.html

  • The whole point of this question was to find a method to backup a running instance without shutting it down. As far as I can tell that is not at all possible with an unconfigured instance that has no repos if you try to do snapshots. My takeaway from this was that it was super important to configure the helm chart we used to have a backing volume. We finally got the resources cleared and before that we just couldn't expect to have our data long term. Our first non-persistant deploy was still useful as a POC. – David West Sep 1 '18 at 17:09
  • As far as I know, you can always configure a destination in ElasticSearch after the fact ... it is an API call. – Evgeny Sep 1 '18 at 22:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.