how to initiate a pgbackrest stanza-create?

[mzwettler2]

we are using s3 storage.

we have to change the whole s3 bucket and repo1-path for some reasons sometimes.

how can we get PGO to initiate a pgbackrest stanza-create if it does not find an existing repo?

it seems that PGO only triggers a pgbackrest stanza-create if there is a completely new configured pgbackrest.repo definition.

[ValClarkson]

Hi @mzwettler2

Sorry you are running into issues. Would you be able to provide me with your manifest and steps? I did some simple testing where I changed buckets and I'm not seeing the same issue.

[mzwettler2]

Hi @ValClarkson

Sorry for being unclear.

When I change (reconfigure) the whole s3 bucket to a new one it works.

If I only change (reconfgure) the repo1-path within the same s3 bucket to a new path it does not work ==>>
ERROR: [103]: unable to find a valid repository

Thanks

[maksonlee]

i changed repo1-path too, and then i got the following error when running scheduled backup,

time="2023-09-24T15:35:12Z" level=info msg="stderr=[ERROR: [055]: unable to load info file '/pgbackrest/kubernetes/awx/backup/db/backup.info' or '/pgbackrest/kubernetes/awx/backup/db/backup.info.copy':\n       FileMissingError: unable to open missing file '/pgbackrest/kubernetes/awx/backup/db/backup.info' for read\n       FileMissingError: unable to open missing file '/pgbackrest/kubernetes/awx/backup/db/backup.info.copy' for read\n       HINT: backup.info cannot be opened and is required to perform a backup.\n       HINT: has a stanza-create been performed?\n]"
time="2023-09-24T15:35:12Z" level=fatal msg="command terminated with exit code 55"

[jmckulk]

Sorry for the delay on this.

I would think that if you are changing the path in the bucket you would want to set it up as a new repo. But the repo-path configuration isn't in the repo-host definition 🤔

@mzwettler2 How are you updating the repo path? Would you be able to provide a before and after spec where you are changing it?

[mzwettler2]

I define my S3 backup location using these CRDs:

PostgresCluster.spec.backups.pgbackrest.repos[index].s3

• bucket
• endpoint
• region

PostgresCluster.spec.backups.pgbackrest

• global.repo[index]-path

If I change the repo-path afterwards, the backup terminates incorrectly. I would have to do a stanza-create first - but that is not possible at the moment.

pgo currently only does a stanza-create when you define a new repo or a new bucket. pgo should also do a stanza-create when the repo-path is changed. At the very least, however, there should be the option of triggering the execution of a new stanza-create with an annotation, for example.

[mzwettler2]

I would need an option to trigger the execution of a new stanza-create.

Anything planned?

[tjmoore4]

Hi @mzwettler2. Thank you for the additional information. Just to verify what you are suggesting, you are effectively just looking for an imperative way to trigger a stanza-create using the configuration already set on the PostgresCluster spec?
Have you tried exec-ing into your Pod and running the stanza create command directly? I'm curious if that method solved your described scenario.

[mzwettler2]

@tjmoore4

you are effectively just looking for an imperative way to trigger a stanza-create using the configuration already set on the PostgresCluster spec?

Yes.

Have you tried exec-ing into your Pod and running the stanza create command directly?

Yes, that should work. The problem is that you would have to find out a lot of specific parameter values beforehand in order to be able to carry out the stanza-create suitable for PGO. this is not practicable for our customers.

[tjmoore4]

@mzwettler2 Understood. I can get something in our backlog for evaluation. In the meantime, I am curious if a command like

kubectl exec -it -n postgres-operator -c database $(kubectl get pods -n postgres-operator --selector='postgres-operator.crunchydata.com/cluster=hippo,postgres-operator.crunchydata.com/role=master' -o name) -- pgbackrest stanza-create --stanza=db

might be useful for this case. This assumes a cluster name of hippo and a namespace named postgres-operator but is otherwise fairly standard in terms of labels and values for lookup.

[mzwettler2]

@tjmoore4

Thanks for adding something to your backlog.
Sorry for beeing unclear. Our customers are not allowed to use kubectl. They must trigger such an action by an annotation.