fdb-kubernetes-operator icon indicating copy to clipboard operation
fdb-kubernetes-operator copied to clipboard

Tester process count results in unhealthy test processes of the Foundationdb

Open iyuroch opened this issue 3 years ago • 5 comments

What happened?

Applying foundationdbcluster with:

processCounts: tester: 4 result in unhealthy tester nodes in foundationdb cluster.

What did you expect to happen?

Have test processes in foundationdb

How can we reproduce it (as minimally and precisely as possible)?

Add processCounts with tester blocks

Anything else we need to know?

Changing class in fdbmonitor configuration from tester to test resolves the issue

FDB Kubernetes operator

0.51.1

Kubernetes version

1.22

Cloud provider

AWS

iyuroch avatar Sep 08 '22 15:09 iyuroch

@johscheuer I'm thinking if we should change definition of process name from Tester to Test or to add this in chunk of fdbmonitor.conf - what do you think would be preferable approach? Short term changing fdbmonitor generation, do you see any issues with changing tester -> test, do we have to take something in consideration with that change?

iyuroch avatar Sep 08 '22 15:09 iyuroch

Changing the name from tester to test make sense to me. To be more precise with the deprecation policies we have to desecrate tester and refer to the newly added test.

Thanks for reporting that issue!

johscheuer avatar Sep 12 '22 08:09 johscheuer

@johscheuer I'll be happy to contribute if you can point me to a direction where to look for CRD's deprecation. Does it also mean that we will not backport this to older CRD's having bug? Maybe we can have combined approach - add implicit rename of class in fdbmonitor generation from "tester" -> "test" and have deprecation policy for CRD?

iyuroch avatar Sep 12 '22 20:09 iyuroch

We have a few examples in our code for how we do the deprecations: https://github.com/FoundationDB/fdb-kubernetes-operator/blob/main/api/v1beta2/foundationdbcluster_types.go#L1669-L1673.We don't maintain the v0 operator release and the best path forward is to upgrade the CRD and the operator to the latest version (there were quite a few bug fixes since then). If you want to make changes to the v0 version of the operator you have to base your changes on the v0 branch.

johscheuer avatar Sep 13 '22 06:09 johscheuer

@johscheuer I've added this -> https://github.com/FoundationDB/fdb-kubernetes-operator/pull/1362 Let me know how to proceed from this point.

iyuroch avatar Sep 26 '22 12:09 iyuroch

Fixed in the PR above. Thanks again!

johscheuer avatar Dec 28 '22 07:12 johscheuer