feat(cluster): allow using existing secret for backup and restore

[itay-grudev]

Closes #197

[itay-grudev]

@phisco I've made your review slightly more complicated as this is the base for my work on adding support for replica clusters and I've fixed and restructured the credential secrets.

[gpothier]

Hi, I merged this branch into main in my fork, as I needed the feature, and I noticed a few issues:

• There are problematic -}} at a few places places in _barman_object_store.tpl in the handling of destinationPath and $secretName; it doesn't work as is, and moreover it complicates things, like having to do {{ " destinationPath: \"s3://" }} instead of just destinationPath: "s3://.
• In _backup.tpl you pass a secretPrefix to barmanObjectStoreConfig, but in _barman_object_store.tpl you expect a secretSuffix.

[itay-grudev]

@gpothier Have you fixed these in your fork? Can I take a look and/or steal your proposal?

[gpothier]

Sure, I just committed it to my fork: https://github.com/gpothier/cnpg-charts/commit/fbdb794f384cfff84f583ce6b1ca62091cb0c2bf

I only fixed the yaml formatting stuff (using }} instead of -}}, as whitespace is part of YAML syntax).

I don't know what your intent is regarding these prefix/suffix things so I did not touch that.

I also added quote to endpointURL as I thought this was part of the problem; it wasn't, but I think it's better to quote in general anyway, and in particular for URLs that usually contain : signs that are not very safe to use in unquoted YAML values. I think it might be a good idea to have a look at other values that are not quoted. See https://helm.sh/docs/chart_template_guide/functions_and_pipelines/:

Let's start with a best practice: When injecting strings from the .Values object into the template, we ought to quote these strings.

[gpothier]

Hi, have you been able to take a look at my suggested fix for this? Do you want me to create another PR with my changes?

[itay-grudev]

I'm super overwhelmed with work right now, so yhea, I would really appreciate any help.

[sxd]

@phisco can you give us a hand here? :D

[itay-grudev]

@Cr4mble Thanks a lot! I would have totally missed that!

[Cr4mble]

@itay-grudev can confirm that the backup to s3 is now working with the current state of the helm chart. The wal files are created in the specific s3 bucket. I used the following values for testing (some are default values):

backups:
  enabled: true
  provider: s3
  secret:
    create: false
    name: "cloudnative-pg-s3-creds"
  s3:
    region: "eu-central-1"
    bucket: "testing-cloudnative-pg-cluster-helm-chart"
    path: ""
  scheduledBackups:
    -
      name: daily-backup
      schedule: "0 0 0 * * *"
      backupOwnerReference: self
      method: barmanObjectStore

[itay-grudev]

@Cr4mble Thanks a lot.

[8ball030]

Thanks a bunch for this guys. Really love the collaboration here :)