node assignment configs are attached on all k8s workloads #1853
Conversation
| @@ -20,9 +20,33 @@ spec: | |||
| resources: | |||
| {{ backup_resource_requirements | to_nice_yaml(indent=2) | indent(width=6, first=False) }} | |||
| {%- endif %} | |||
| {% if db_management_pod_node_selector %} | |||
| {% if task_node_selector %} | |||
There was a problem hiding this comment.
@ranvit The task_node_selector parameter only exists in the installer role (AWX CR), not in the back role (AWXBackup CR).
So db_management_pod_node_selector is correct.
If you need topologySpreadConstraints, tolerations, and affinity as well, those params would need to be added to the AWXBackup CR and should be prefixed with db_management_pod_*
There was a problem hiding this comment.
This PR also needs a rebase now to resolve conflicts.
|
Hello @ranvit, please let us know if you are still interested in merging this PR. If so, please be sure to address the comments left by @rooftopcellist and resolve any merge conflicts. Thank you for your time! |
|
SUMMARY
The db migration pod and the backup process pod have failed to schedule on a k8s cluster where every node has a taint. So this PR adds the task_* related node-assignment params and the global node-assignment params as fallback to every k8s workload that didn't have node-assignment configuration available on them.
This fixes issues like #1774
ISSUE TYPE
ADDITIONAL INFORMATION
I PR'd a similar fix with #1804, only targeting the migration pod. Realised I need somethign similar on the backup pod.
The task_* params seem well suited for most of the k8s workloads except for the mesh-ingress deployment, so I had that only use the global params