Thitiwat Jumpha:
Hi, I’m in a scheduler exam, I try to create a custom scheduler when Pod has created the state of Pod is CrashLoopBackOff .
Raamkanna Saranathan:
seems a lot of people are complaining about this. it “magically” worked for me the first time a couple days ago but when i tried again, it failed complaining about some volume issue. seems like unstable lab.
make sure you do the following:
• leader-elect: false
• address=0.0.0.0
• port: anything other than 10250-2 - see here for details on already used ports: https://kubernetes.io/docs/setup/production-environment/tools/kubeadm/install-kubeadm/#check-required-ports
then you should be good. if it fails despite that, try kubectl logs <your-scheduler> and kubectl events -n kube-system to debug and/or restart lab
Mohamed Ayman:
We need to set the --leader-elect=false here, because we don’t want to select a leader between the two existing schedulers but having both function at the same time. The leader election is used when multiple master nodes are setup and with that multiple schedulers
Ejay Tumacder:
Check the Solution tab, there are details there that aren’t in the lecture.
• Set --leader-elect=false
• Use a different port not used by controlplane, e.g. 10282. Also set this port in the liveness and startup probes. You can check what ports aren’t used with sudo netstat -tulpn | grep LISTEN
• Set --scheduler-name=my-scheduler
• --secure-port=0 then set schemes for the probes to HTTP
Don’t really understand the secure-port part yet. Hope this makes more sense as the course progresses.