Open tooptoop4 opened 1 month ago
Have you tried PodSpecPatch
?
PodSpecPatch adds unnecessary latency
Argo is not inherently a low-latency task scheduling system, so the latency is almost negligible compared to the scheduling time.
Moreover, to implement this feature, the method is almost identical to PodSpecPatch
.
https://argo-workflows.readthedocs.io/en/release-3.5/cost-optimisation/#configure-executor-resource-requests shows how to set default limits (ie cpu/memory) for all workflows
would like ability for one workflow to do an override of setting no limits i tried
cpu: null
but it did not work