Standardise CPU and Memory Resources for Argo Workflows#1322
Conversation
|
|
||
| workflowDefaults: | ||
| spec: | ||
| podSpecPatch: | |
There was a problem hiding this comment.
I think that this modifies the default Workflow pods resources? What is the motivation for this?
I don't think this is needed for AP-1119?
There was a problem hiding this comment.
You are correct — this change goes beyond AP-1119.
The intention was to align workflow execution pods with the same resource defaults, but this is not strictly required for this ticket.
I can remove workflowDefaults.spec.podSpecPatch from this PR and limit the change to controller/server resources only if preferred, and raise a separate PR for workflow pod defaults.
There was a problem hiding this comment.
Setting default limits on pods may be a good idea but we should discuss what they should be before implementing.
fix: bump workflows chart version fix: update workflow controller cpu and memory resources fix: update workflow controller cpu and memory resources
hazdl
force-pushed
the
haz/cpu-memory-limits
branch
from
c4bc8d0 to
37ddd8f
Compare
|
|
||
| workflowDefaults: | ||
| spec: | ||
| podSpecPatch: | |
There was a problem hiding this comment.
Setting default limits on pods may be a good idea but we should discuss what they should be before implementing.
2 participants
This PR standardises CPU and memory requests/limits across the Argo Workflows Helm chart to ensure consistent resource allocation for controller, server, and workflow execution pods.
-Updated controller resources to: requests: 500m CPU / 2Gi memory
limits: 1 CPU / 4Gi memory
-Updated server resources to: requests: 1 CPU / 1Gi memory
limits: 2 CPU / 2Gi memory
-Added/updated workflowDefaults.spec.podSpecPatch to enforce workflow pod resources:
requests: 500m CPU / 2Gi memory
limits: 1 CPU / 4Gi memory
-Applies to: new workflow executions only
No changes required to existing workflows
Verification:
Verified via helm template rendering that workflowDefaults.spec.podSpecPatch applies CPU and memory requests/limits to the workflow main container. Confirmed in rendered output:
workflowDefaults.spec.podSpecPatch → containers.main.resources
Confirmed using Helm-rendered manifests and grep validation.
grep -n "podSpecPatch" rendered.yaml
grep -n "2Gi" rendered.yaml
grep -n "500m" rendered.yaml