Resource Limits vs Requests

[jonstelly]

There seems to be some sort of requirement that resource Limits <= Requests for all containers in a pod? Another report of this from 2019: #17.

I've requested some documentation improvements MicrosoftDocs/azure-docs#78032 but is there any way to improve this? For my case I've got linkerd and knative sidecars being injected into my pod and if I set limits and requests equal on those it eats up all the resources in my non-ACI nodes.

Would it be possible to change the logic so ACI calculates the sum of all container's limits and requests, then uses the maximum value?

time="2021-07-07T20:33:17Z" level=error msg="failed to create container group default-my-worker-00007-deployment-68f476fd9b-phjzn" azure.region=... azure.resourceGroup=me-dev-...-kube error="api call to https://management.azure.com/subscriptions/...: got HTTP response status code 400 error code "ContainerLimitGreaterThanContainerGroupTotalRequest": The 'Cpu' limit '1.5' in container 'me' is greater than the total request '0.3' of container group 'default-my-worker-00007-deployment-68f476fd9b-phjzn'." key=default/my-worker-00007-deployment-68f476fd9b-phjzn method=aci.createContainerGroup name=my-worker-00007-deployment-68f476fd9b-phjzn namespace=default node=virtual-node-aci-linux operatingSystem=Linux phase=Pending pod=my-worker-00007-deployment-68f476fd9b-phjzn provider=azure reason=ProviderFailed uid=e2ddabce-1032-4768-80ac-110720d38c8f watchedNamespace= workerId=12

[helayoty]

@mainaksarcar @fnuarnav, wdyt?

[joanna-jasnowska-wttech]

hello, is there any update on this?