Closed ghost closed 5 years ago
So generally, when heketi is managing dedicated gluster nodes rather than gluster-in-containers (from here called an "independent" cluster) heketi must be configured to update /etc/fstab rather than /var/lib/heketi/fstab. Only the container based systems have scripts that operate on /var/lib/heketi/fstab. However, if you chose to use /var/lib/heketi/fstab for a reason, like you don't want heketi changing /etc/fstab you could create your own "init script" to run 'mount -a --fstab /var/lib/heketi/fstab'. You'd need to have this run before glusterd starts.
Ok, fine for me to create my own init script. Thanks for the quick answer!
Maybe not the right place to ask, but let me try :-) From your answer, I can understand that it's possible to choose which file Heketi will write its own mount: may you please address me where this configuration can be done? Would it be possible to do it also when Heketi runs as a Pod?
I was assuming that you installed using gk-deploy. If not let me know. When you use it it comes with a template config file for heketi. If you modify this file it will take your modifications. However, if you do not specify "fstab" in the config anywhere it should automatically use /etc/fstab when you use the ssh executor. How heketi itself is run, container or not, doesn't matter. It's fully controlled by the config file and the environment vars.
No, I'm not using gk-deploy. I installed RHGS using yum, then I followed the guide that can be found here: https://github.com/heketi/heketi/blob/46a8eaa3bfbc488b35ebaa27ff7ca12a594fe549/docs/admin/install-kubernetes.md, skipping the GlusterFS installation part. I guess that the file that Heketi is using to write its own mounts is part of the Heketi configuration, isn't it? Can you tell me were can I find it?
Thanks again for your support!
I think I've got it:
Environment:
HEKETI_DB_PATH: /var/lib/heketi/heketi.db
HEKETI_FSTAB: /var/lib/heketi/fstab
HEKETI_SNAPSHOT_LIMIT: 14
I don't know if I can close this issue by myself, or if you should do.
Thanks again!
Hello, I have an issue with GlusterFS (Red Hat Gluster Server) cluster running outside of Kubernetes, but managed by an Heketi instance running as a pod.
My configuration is:
My RHGS cluster is running outside of the K8s cluster, in a dedicated environment. Heketi is instead running as a Pod in the K8s cluster.
After restarting one (actually, all) the GlusterFS nodes, no volume cames up, because the
/var/lib/heketi/fstab
mounts have not been performed automatically. This leads me to the question: who is supposed to configure the GlusterFS nodes to perform those mounts automatically? I took for granted that it was part of the Heketi/GlusterFS setup, but it seems not. Or it is, and I have to look to some specific service which has not been started correctly?Thanks!