RedHatGov / service-mesh-workshop-dashboard

Materials for a workshop on OpenShift Service Mesh
26 stars 23 forks source link

RHPDS resources (POD count per node > maxPods) are not enough #21

Closed tgubeli closed 2 years ago

tgubeli commented 2 years ago

I've setup this workshop for 40 participants. So 40 service mesh deployments. Openshift by default has a limit of pods per worker node (250). This limit was met very quickly (the ocp "large" cluster instance has only 2 workers nodes). That caused that no more applications could be deployed in the cluster.

You must consider the amount of pods in the deployment script and relate that with the amount of participants (the amount of pods per service mesh instances and the amount of pods that the jupyter spawner creates). Just changing the maxpods per worker node parameter to 500 (which is supported , the default value is 250 pods per worker node) or adding more worker nodes to the openshift deployment/cluster. Ref: https://docs.openshift.com/container-platform/4.10/scalability_and_performance/planning-your-environment-according-to-object-maximums.html

andykrohg commented 2 years ago

Hey @tgubeli thanks for the heads up. We've engaged the RHPDS team to help get this resolved. I'll report back here when we have a fix in place. As a workaround, you could manually increase the number of replicas on your worker node MachineSet after the cluster is provisioned

andykrohg commented 2 years ago

I just deployed a 40-user workshop using a modified user-to-worker-node ratio and there are 8 worker nodes - hopefully that should suffice! 😄 Thanks again for bringing this to our attention @tgubeli

tgubeli commented 2 years ago

I just deployed a 40-user workshop using a modified user-to-worker-node ratio and there are 8 worker nodes - hopefully that should suffice! 😄 Thanks again for bringing this to our attention @tgubeli

Hey Andy, no problem at all. Thank you!