Move pgbackrest-restore test to Kyverno Chainsaw #4228
Conversation
This test has a number of scripts and jobs that pass and share data. Chainsaw's bindings and templates are a nice way to break this test up, and its "catch" operations provide good context when a step fails. Tested with Kyverno Chainsaw 0.2.12 See: https://kyverno.github.io/chainsaw/main
This helped me reproduce a race in "pg_ctl start" during Postgres recovery. Issue: PGO-1945
| use: { template: '21-lose-data.yaml' } | ||
|
|
||
| - name: 'Point-In-Time Restore' | ||
| use: { template: '22-point-in-time-restore.yaml' } |
There was a problem hiding this comment.
Is it worth it to regularize the file/template names? I always liked the numbered files because you can just read through a directory (well, could, before we started to use the subdir files), but if Chainsaw uses this file to sequence the tests, maybe another convention is worth talking about (at a later date!)
There was a problem hiding this comment.
Definitely. I think I started with numbers close to the KUTTL ones to ease review/migration, but where I landed is not that!
I like the filenames going in a similar order as the test, too, but now they can get reused so... 🤷🏻
Maybe something more like "scenario one," "... two," 🤔 🌱
There was a problem hiding this comment.
Yeah, if the templates are being reused (hopefully!), then putting them in order doesn't make sense. Maybe descriptive titles could make it easier to read?
| run: | | ||
| curl -Lo /usr/local/bin/kubectl-kuttl https://github.com/kudobuilder/kuttl/releases/download/v0.13.0/kubectl-kuttl_0.13.0_linux_x86_64 | ||
| chmod +x /usr/local/bin/kubectl-kuttl | ||
|
|
There was a problem hiding this comment.
We don't need to install KUTTL any longer?
There was a problem hiding this comment.
The Make targets do go run ...@latest by default. The binary download is faster, but not as easy to latest.
I'm on the fence. Do you have a preference?
| Confirm that data was restored to the point-in-time and the cluster is healthy | ||
| use: { template: '23-verify-primary.yaml' } | ||
|
|
||
| - name: 'Confirm Replica' |
There was a problem hiding this comment.
Again: not for today, but I wonder if we could templatize a verify step and reuse it in our 4 verify steps. Note: this is the longest sustained look I've had at chainsaw
There was a problem hiding this comment.
Hmm. I thought I was reusing verify 02 and 12, but i see each only once.
Should those be like this?
- Create Data [A] on [one]
- Verify Data [A] on [one] <-- new
- Clone [one] to [two]
- Verify Data [A] on [two]
| @@ -0,0 +1,62 @@ | |||
| apiVersion: chainsaw.kyverno.io/v1alpha1 | |||
There was a problem hiding this comment.
Is this file getting called?
There was a problem hiding this comment.
It is not. I used it while diagnosing a bug about parameters in the WAL stream a long time ago.
I called this multiple times from the main test to... 🤔 raise and lower max_connections maybe. 44b1847
There was a problem hiding this comment.
🔧 If the file stays, it deserves a comment explaining what it does and why it isn't currently used.
There was a problem hiding this comment.
This is looking a lot like 11 ...
Checklist:
Type of Changes:
Other Information:
This test failed in this other PR, and the KUTTL logs did not give me enough information to diagnose the cause. Chainsaw did much better and unblocked me over there.