[OVNDBCluster] Check bound error on reconcileService #297
Conversation
|
Skipping CI for Draft Pull Request. |
|
Build failed (check pipeline). Post https://review.rdoproject.org/zuul/buildset/11f304f5a4a741fdb4a8e825bbe785dc ❌ openstack-k8s-operators-content-provider FAILURE in 8m 22s |
|
|
||
| package common | ||
|
|
||
| // Helper function while can't use buildin min |
|
/lgtm |
| @@ -749,6 +751,12 @@ func (r *OVNDBClusterReconciler) reconcileServices( | |||
| if err != nil { | |||
| return ctrl.Result{}, err | |||
| } | |||
| // It can be possible that not all pods are ready, so DNSData won't | |||
There was a problem hiding this comment.
Trying to understand why can't we just move this length check below at L723 and avoid/remove the minLenth calculator?
There was a problem hiding this comment.
You mean "above"?
I suggested to still attempt to create DNS records for running pods, even if not all of them are running, yet.
The fact you ask suggests this should be documented in a comment.
There was a problem hiding this comment.
Yes sorry, i meant "above". Not sure how much that helps to create DNSData for partial ip list. But just wanted to understand the reasoning for that, as this would just lead to unnecessary respawns for DNSMasq pod and if could avoid the hacks around Min checks. But if we still want to go with this i am ok
There was a problem hiding this comment.
So, one scenario that comes to my mind is where there are not enough PVC available for all the ovndb replicas (5 replicas for example, only 3 PV available). With the current approach cluster will work as 3 replicas will be available.
If we just return whenever status.Replicas > len(podList) in the case mentioned above any ovsdbserver would be accessible using the CNAME.
Need to look regarding respawning DNSMasq.
There was a problem hiding this comment.
It will respawn once new pods show up, yes. That's sadly how dnsmasq is implemented right now.
There was a problem hiding this comment.
Yeah, I was not sure about the behavior regarding the reconcile when the DNSData doesn't change.
My concern was in the scenario where replicas > len(podList), that the reconcile will be triggered until condition is fulfilled, and in every reconcile loop (with the current code) the DNSData creation will be called, and I was afraid that it would trigger the recreation of the dnsmasq pod everytime. But that only will happen if the DNSData has different information.
I don't think there's much of an overhead regarding not waiting until all pods are ready. Having said that, I'm okay with any of the two solutions.
WDYT?
There was a problem hiding this comment.
I think everyone involved is ok with either way. :) This is not very consequential. But if you keep creating DNSData while not all pods are ready, please add a comment explaining the thinking behind this.
|
/lgtm |
|
lgtm, can you rebase it on main branch |
Related-Issue: OSPRH-6798
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: averdagu, karelyatin The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
openshift-merge-bot
bot
merged commit d45e9a0
into
openstack-k8s-operators:main
Related-Issue: OSPRH-6798