Fix bug in ComputeDomain object being written to mutation cache #721
Conversation
github-project-automation
bot
moved this to Backlog
in Planning Board: k8s-dra-driver-gpu
Signed-off-by: Kevin Klues <[email protected]>
klueska
force-pushed
the
fix-newcd-bug
branch
from
48b41ad to
ecd1393
Compare
|
/ok to test ecd1393 |
| newCD.Status.Nodes = updatedNodes | ||
| if _, err := m.config.clientsets.Nvidia.ResourceV1beta1().ComputeDomains(newCD.Namespace).UpdateStatus(ctx, newCD, metav1.UpdateOptions{}); err != nil { | ||
| newCD, err = m.config.clientsets.Nvidia.ResourceV1beta1().ComputeDomains(newCD.Namespace).UpdateStatus(ctx, newCD, metav1.UpdateOptions{}) | ||
| if err != nil { |
There was a problem hiding this comment.
Connecting dots: we noticed this before; that's the only unresolved comment on #651:
jgehrcke
left a comment
jgehrcke
left a comment
There was a problem hiding this comment.
Thank you! Of course this is what we want, for various reasons.
However, I still need to convince myself if this patch alone fixes a well-understood bug.
I am slowly catching up with the Slack thread that sparked reviewing this code section (and making the change).
Here, we make a change in removeNodeFromComputeDomain(). I wonder how that part of the business logic relates to the scenario in the user's environment. It appears as if that scenario is about cleanly growing a CD over many nodes (and hence node removal from that CD should not be part of that scenario).
In any case, let's get this in. And then worry about classification of this patch later when we have learned/understood more.
|
Not sure about the bug this fixes, but let's certainly land this. |
github-project-automation
bot
moved this from Backlog
to Closed
in Planning Board: k8s-dra-driver-gpu
|
/cherry-pick release-25.8 |
|
🤖 Backport PR created for |
2 participants
No description provided.