K
Q

Kubernetes scheduler: watch of *api.Pod ended with error: unexpected end of JSON input

3/12/2015

Yesterday service worked fine. But today when i checked service's state i saw:

Mar 11 14:03:16 coreos-1 systemd[1]: scheduler.service: main process exited, code=exited, status=2/INVALIDARGUMENT
Mar 11 14:03:16 coreos-1 systemd[1]: Unit scheduler.service entered failed state.
Mar 11 14:03:16 coreos-1 systemd[1]: scheduler.service failed.
Mar 11 14:03:16 coreos-1 systemd[1]: Starting Kubernetes Scheduler...
Mar 11 14:03:16 coreos-1 systemd[1]: Started Kubernetes Scheduler.
Mar 11 14:08:16 coreos-1 kube-scheduler[4659]: E0311 14:08:16.808349    4659 reflector.go:118] watch of *api.Service ended with error: very short watch
Mar 11 14:08:16 coreos-1 kube-scheduler[4659]: E0311 14:08:16.811434    4659 reflector.go:118] watch of *api.Pod ended with error: unexpected end of JSON input
Mar 11 14:08:16 coreos-1 kube-scheduler[4659]: E0311 14:08:16.847595    4659 reflector.go:118] watch of *api.Pod ended with error: unexpected end of JSON input

It's really confused 'cause etcd, flannel and apiserver work fine.

Only some strange logs are for etcd:

Mar 11 20:22:21 coreos-1 etcd[472]: [etcd] Mar 11 20:22:21.572 INFO      | aba44aa0670b4b2e8437c03a0286d779: warning: heartbeat time out peer="6f4934635b6b4291bf29763add9bf4c7" missed=1 backoff="2s"
Mar 11 20:22:48 coreos-1 etcd[472]: [etcd] Mar 11 20:22:48.269 INFO      | aba44aa0670b4b2e8437c03a0286d779: warning: heartbeat time out peer="6f4934635b6b4291bf29763add9bf4c7" missed=1 backoff="2s"
Mar 11 20:48:12 coreos-1 etcd[472]: [etcd] Mar 11 20:48:12.070 INFO      | aba44aa0670b4b2e8437c03a0286d779: warning: heartbeat time out peer="6f4934635b6b4291bf29763add9bf4c7" missed=1 backoff="2s"

So, I'm really stuck and don't know what's wrong. How can i resolve this problem? Or, how can i check details log for scheduler. journalctl give me same logs like systemd status

-- Suvitruf - Andrei Apanasik
etcd
kubernetes

Similar Questions

Creating isolated Docker Zones
gcloud kubectl failing to connect to Kubernetes API Endpoint
Start kubernetes container with specific command
Does the kubernetes scheduler support anti-affinity?

1 Answer

3/12/2015

Please see: https://github.com/GoogleCloudPlatform/kubernetes/issues/5311

It means apiserver accepted the watch request but then immediately terminated the connection.

If you see it occasionally, it implies a transient error and is not alarming. If you see it repeatedly, it implies that apiserver (or etcd) is sick.

Is something actually not working for you?

-- lavalamp
Source: StackOverflow