search

camunda-consulting / c8-devops-workshop

DevOps Workshop for Camunda 8
0 stars 1 forks source link

Lab 3 / restoration #16

Closed pierre-yves-monnet closed 8 months ago

pierre-yves-monnet commented 9 months ago

having a Lab to back up and never playing the restoration is too bad, especialy because there is no REST API to restore Zeebe, but it has to be done via a command. 1/ we should start the cluster with 

  clusterSize: 3
  partitionCount: 3
  replicationFactor: 3

to be more realistic

There is an excellent document started by: DRI: Stefan Wiese | Contributors: Tobias Stegemann David Paroulek Falko Menge Christophe Dame Paul Lungu

It shows how to configure the backup config for Zeebe and how to restore it. It does not show the complete backup procedure for the whole product stack!

(sorry, I didn't retrieve the URL, we should ask Stefan)

It will be nice to create a step, because the cluster is created by Helm: how to restore then? Create a custom Deployment, maybe, and start it? This deployment may just contain the restore part (no need to have an init container).

j-lindner commented 9 months ago

The mentioned excellent doc: https://docs.google.com/document/d/16rHtbavxJU7KuyBjgT0_NPQvieAiReFfAbxRW8Nb9gs/edit#heading=h.ct4eck1q9lyq

jonathanlukas commented 8 months ago

Done via https://github.com/camunda-consulting/c8-devops-workshop/commit/4e65ba6e595b6387a639232cae6dd21f2f6524cb

ManuelDittmar commented 8 months ago

Hi @jonathanlukas thanks for taking the initiative. To be honest I don't agree with having a setup with 3 brokers. For a lot of participants this means they cannot perform the exercise because they run on a local machine with one node. Of course you could also adjust the pod affinity, but I would prefer to keep the exercises as simple as possible. Also for implementing new features / exercises in the training I think we should establish a review process and not just merge into the main branch.

jonathanlukas commented 8 months ago

Hi @ManuelDittmar I agree with both proposals.

Therefore, I will set the cluster settings to 111 for this exercise.

For future FR implementations, I will use PRs