Hello all,
I am trying to install a ceph environment for testing, focused on block devices for virtualization and object storage. I am facing some issues.
My environment consists of one server running ceph-deploy and mon; 3 cluster nodes with 3 disks each running a total of 3 OSD's; and one object gateway. All servers are virtualized, running ubuntu 13.04 on XenServer hosts. When installing, I created a cluster with non-default name.
The issues:
1) I couldn't create the OSD with non-default filesystem. I tried using "--fs-type btrfs ", as stated here. The option was not accepted. How can I use btrfs using ceph-deploy?
A bug was just opened for this inthe tracker. However, I believe you can work around it be editing the ceph.conf (before pushing it out to the OSDs) and specifying "fs type = btrfs" in the OSD section.
2) When trying to restart a OSD - in my case, to apply a changed configuration for cluster network, the command "sudo /etc/init.d/ceph -c <cluster>.conf stop osd.<num>" is accepted, but doesn't shut down the osd. I had to kill the process to achieve it, and run the command line request to run it again.
On Ubuntu you should be using Upstart to manage the processes. "sudo stop ceph-osd id=<num>", iirc.
Btw, I still have a number of connections between OSDs across the public network, even after configuring the cluster network.
Hmm, I think that's so they can do pings to measure the health of both interfaces.
I'm not sure about the gateway; somebody else will have to pick that up! :)
-Greg
3) I'm also facing problems running radosgw - when running /etc/init.d/radosgw start, the request is accepted, but no process exists; but I'm still gathering more information and double checking the object gateway installation.
Any help would be welcome.
Regards
Gerd Jakobovitsch
--
Software Engineer #42 @ http://inktank.com | http://ceph.com
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
