Fixed re-ip error when restart the cluster #856
Conversation
cchen-vertica
force-pushed
the
cchen/fix-openshift-issues
branch
from
ddd9ef8
to
18b146c
Compare
cchen-vertica
force-pushed
the
cchen/fix-openshift-issues
branch
from
18b146c
to
adac04c
Compare
| @@ -389,6 +392,7 @@ func (p *PodFacts) collectPodByStsIndex(ctx context.Context, vdb *vapi.VerticaDB | |||
| return err | |||
| } | |||
| pf.hasNMASidecar = vk8s.HasNMAContainer(&pod.Spec) | |||
| pf.isNMAContainerReady = vk8s.IsNMAContainerReady(pod) | |||
There was a problem hiding this comment.
What about when nma is not running in a separate container?
There was a problem hiding this comment.
This will return false. Now NMA is always running as a sidecar so I didn't consider that case.
If we need to cover the case nma is not running in a seprate container, we need to update gatherScript to check if nma process exists. However, that is for the old vertica versions, and it isn't worth doing that because the ReIP error is rare and don't affect anything.
| canReIPAllDownPods := containPods(reIPPods, downPods) | ||
| if !canReIPAllDownPods { | ||
| r.Log.Info("Not all restartable pods are qualified to re-ip. Need to requeue restart reconciler") | ||
| return ctrl.Result{Requeue: true}, nil | ||
| } |
There was a problem hiding this comment.
Can you explain this part?
|
@cchen-vertica I can't log in to my PC. I am asked to provide BitLocker recovery key. Can you reach out to me through linkedin or my personal email(paulin.nguetsop@yahoo.com). |
Pinged you through LinkedIn. |
| @@ -278,6 +297,11 @@ func (r *OnlineUpgradeReconciler) postCreateNewSubclustersMsg(ctx context.Contex | |||
| // exists in the main cluster. This is a pre-step to setting up replica group B, which will | |||
| // eventually exist in its own sandbox. | |||
| func (r *OnlineUpgradeReconciler) assignSubclustersToReplicaGroupB(ctx context.Context) (ctrl.Result, error) { | |||
| // If we have already promoted sandbox to main, we don't need to do this step | |||
| if vmeta.GetOnlineUpgradeSandboxPromoted(r.VDB.Annotations) == vmeta.SandboxPromotedTrue { | |||
There was a problem hiding this comment.
Move this to vdb: vdb.IsOnlineUpgradeSandboxPromoted().
| } | ||
| return 0 | ||
| }) | ||
| return restartablePrimaryNodeCount >= (primaryNodeCount+1)/2 |
There was a problem hiding this comment.
restartablePrimaryNodeCount > primaryNodeCount/2 looks simpler.
There was a problem hiding this comment.
restartablePrimaryNodeCount > primaryNodeCount/2 cannot work well. If restartablePrimaryNodeCount=2 and primaryNodeCount=4, we will return false which is wrong.
This PR fixed re-ip error when restart the cluster by checking all pods are up and NMA is running in all pods. Without the check, vclusterOps re-ip could fail when NMA is not running or half of the primary nodes are not up.
This PR fixed re-ip error when restart the cluster by checking all pods are up and NMA is running in all pods. Without the check, vclusterOps re-ip could fail when NMA is not running or half of the primary nodes are not up.
This PR fixed re-ip error when restart the cluster by checking all pods are up and NMA is running in all pods. Without the check, vclusterOps re-ip could fail when NMA is not running or half of the primary nodes are not up.
This PR fixed re-ip error when restart the cluster by checking all pods are up and NMA is running in all pods. Without the check, vclusterOps re-ip could fail when NMA is not running or half of the primary nodes are not up.
This PR fixed re-ip error when restart the cluster by checking all pods are up and NMA is running in all pods. Without the check, vclusterOps re-ip could fail when NMA is not running or half of the primary nodes are not up.
This PR fixed re-ip error when restart the cluster by checking all pods are up and NMA is running in all pods. Without the check, vclusterOps re-ip could fail when NMA is not running or half of the primary nodes are not up.