Single node "best-effort drain" during upgrade #3445
Labels
kind:bug
Something isn't working
release:blocker
An issue that blocks a release until resolved
topic:lifecycle
Issues related to upgrade or downgrade of MetalK8s
Milestone
Comments
TeddyAndrieux
added
kind:bug
TeddyAndrieux
added a commit
that referenced
this issue
Jul 13, 2021
When running in single node cluster we want to drain the node so that we have as less pod running on the node as possible. Fixes: #3445
TeddyAndrieux
added a commit
that referenced
this issue
Jul 13, 2021
When running in single node cluster we want to drain the node so that we have as less pod running on the node as possible. Fixes: #3445
TeddyAndrieux
added a commit
that referenced
this issue
Jul 13, 2021
When running in single node cluster we want to drain the node so that we have as less pod running on the node as possible. Fixes: #3445
TeddyAndrieux
added a commit
that referenced
this issue
Jul 13, 2021
When running in single node cluster we want to drain the node so that we have as less pod running on the node as possible. Fixes: #3445
TeddyAndrieux
added a commit
that referenced
this issue
Jul 13, 2021
When running in single node cluster we want to drain the node so that we have as less pod running on the node as possible. Fixes: #3445
TeddyAndrieux
added a commit
that referenced
this issue
Jul 13, 2021
When running in single node cluster we want to drain the node so that we have as less pod running on the node as possible. Fixes: #3445
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Component:
'lifecycle'
What happened:
Due to a "bug" (:question:) in kubelet when upgrading a single node cluster with a bunch of Pods running, upgrade fail because it takes too much time to restart static Pods (like apiserver).
See: kubernetes/kubernetes#103658
What was expected:
Single node upgrade to work properly
Resolution proposal (optional):
In order to avoid that kind of issue let's drain the node during the upgrade even if it's not really needed since it's .... single-node cluster so anyway every services may/will have downtime during the upgrade process.
Draining may not be possible on a single node as you do not have any other node to schedule needed Pods (e.g.: If you have some PodDisruptionBudget configured in the cluster).
Add a "best-effort" drain used for single-node drain during the upgrade, this drain just cordon the node and evict all Pods possible, and do not retry/fail if one Pod cannot be evicted just continue the "classic upgrade process"
The text was updated successfully, but these errors were encountered: