On Thu, 2012-03-22 at 15:43 -0400, Hans Lellelid wrote:
> I'm probably missing something, but it's unclear to me how to create > this local storage mountpoint on that host. (I don't see anything in > the Node TUI.) I tried simply specifying "/data/images" in the hopes > that it was automatically created, but the Admin console just spun > indefinitely (eventually refreshing the page revealed that the store > was note created). > > > I'm currently rebuilding that Node as a F16+VDSM box instead -- so I > can get in there and set this up manually (is there a way to get into > schell on oVirt Nodes and mess with the [F16-based?] OS?)
You can drop to the shell using F2 and create the directory that way.
Thanks, that's really helpful. I'm sure it was mentioned somewhere, but I was trying different consoles with "root" etc. and unsure on how to get to the terminal.
IMO, since we restrict the location you can create a local data store, we should automatically create that when configuring the local datastore.
We can't actually pre-create the directory since it's a mounted filesystem, but it might be possible to do it at installation or boot time. The right place is in vdsm when it's trying to create the local data store.
So, after initially attempting to just "create" the directory and being told it was a read-only filesystem, I did a little more investigation and realized that the "Data" volume (in /dev/HostVG) was simply not mounted. I mounted this with the command:
mount /dev/HostVG/Data /data
And now I have a /data/images directory (as well as /data/core). I'm not entirely sure why this wasn't mounted since I see it in /etc/fstab. In fact, when I discovered it was in /etc/fstab, I tried just
mount /data
And that also worked just fine. So puzzled that thsi was not mounted when the system booted up. Maybe I missed an error message, but dmesg isn't saying anything obvious (not sure if that's where I'd see mount errors or not). Maybe this is a known issue.
I had heard this from a couple people previously, but it always seemed to present as a one off problem. We've managed to reproduce it consistently, so it seems that it's a legitimate problem that we need to fix. Thanks Mike
Thanks again for the help -- this definitely got me moving in the right direction. Hopefully I'm moments away from being able to get that first VM running.
Thanks, Hans