0

currently I am following the k8s guide "from scratch" (found here: LINK, up to this point everything seems pretty clear, though I start without doing all that certificate stuff.

I started kubelet on master and added the etcd.manifest file found here

Kubelet finds the manifest file and start two containers: One container running etcd and on "pause" handling the port mapping: 0.0.0.0:2379-2380->2379-2380/tcp.

Unfortunately, I cannot access etcd by doing a telnet IP_OF_SERVER:2379 oder 2380. Doing this on localhost, (telnet 127.0.0.1 2379) results in a connection that gets closed directly.

After a couple of seconds, kubeletes health check fails and the container gets shut down: 

Jul 05 16:32:43 reactprd01.abc.com kubelet[18003]: I0705 16:32:43.834839   18003 manager.go:1751] pod "etcd-serverreact-reactprd01.abc.com_kube-system(b5a9f2d68d0d6160eed4a802d1afa5a7)" container "etcd-container" is unhealthy, it will be killed and re-created.
Jul 05 16:32:43 reactprd01.abc.com kubelet[18003]: I0705 16:32:43.835232   18003 manager.go:1371] Killing container "02df58ce685fd220cc19e6a38e8a5823de2699d869b5bb1789c50f1a6baa462b etcd-container kube-system/etcd-serverreact-reactprd01.abc.com" with 30 second grace period

Any idea on whats going on? I replaced all placeholder (e.g. {{ port }} with data). I could not find anything on google as well.

Thanks for any advice :)

Jannick
  • 131
  • 3

1 Answers1

0

According to the etcd.manifest file you linked to, etcd's stdout and stderr are redirected to a file mounted as a volume from /var/log/etcd-serverreact.log. It's possible that the file doesn't exist, or that your kernel doesn't support bind-mounting files (as opposed to directories) into Docker containers. And, for the long shot, maybe there's an error message in the file.

aecolley
  • 943
  • 4
  • 15