The tool currently is out of CPU limits and pods too:

limits.cpu              2       2
pods                    4       4

How much more will you need on those? Do you imagine you will end up hitting any other limits (kubectl describe quota) in the near future?

I’ve lowered the CPU requests/limits of the non-webservice deployments, now we’re at 1 out of 2 CPUs (500m webservice, 200m+200m+100m the other three). I think the CPU quota should be alright for now.

The pod limit is a good point – I definitely need at least one more pod for the function evaluator (for five pods total: webservice, pygments-server, function-orchestrator, function-evaluator, update). And if Kubernetes likes to create new pods before removing old ones (e.g. when I try to kubectl apply a config change), then it would probably be useful to have a slightly higher limit – maybe 7?

Mentioned in SAL (#wikimedia-cloud) [2021-05-29T14:57:17Z] <lucaswerkmeister> lowered CPU requests/limits of non-webservice deployments (cf. T283970#7123072)

+1 looks good to me

Increased the pods quota to 7:

dcaro@tools-sgebastion-07:~$ kubectl describe resourcequota tool-notwikilambda --namespace tool-notwikilambda
Name:                   tool-notwikilambda
Namespace:              tool-notwikilambda
Resource                Used    Hard
--------                ----    ----
configmaps              1       10
limits.cpu              1       2
limits.memory           4608Mi  8Gi
persistentvolumeclaims  0       3
pods                    4       7
replicationcontrollers  0       1
requests.cpu            650m    2
requests.memory         3840Mi  6Gi
secrets                 1       10
services                3       3
services.nodeports      0       0

Enjoy! Let me know if there's any issues.

Alright, thanks :)