----- Original Message -----
> From: "Steve Kilduff" <kild...@gmail.com>
> To: users@ovirt.org
> Sent: Tuesday, August 25, 2015 6:27:26 PM
> Subject: Re: [ovirt-users] Cannot add Posix Storage
> 
> Same on centos6 trying to use a bind mount:
> 
> Thread-5189::ERROR::2015-08-25
> 15:22:26,318::hsm::2379::Storage.HSM::(connectStorageServer) Could not
> connect to storageServer
> Traceback (most recent call last):
> File "/usr/share/vdsm/storage/hsm.py", line 2376, in connectStorageServer
> conObj.connect()
> File "/usr/share/vdsm/storage/storageServer.py", line 220, in connect
> self.getMountObj().getRecord().fs_file)
> File "/usr/share/vdsm/storage/mount.py", line 271, in getRecord
> (self.fs_spec, self.fs_file))
> OSError: [Errno 2] Mount of `/data/test` at
> `/rhev/data-center/mnt/_data_test_` does not exist
> 
> And the mount gets mounted on the OS:
> 
> /data/test on /rhev/data-center/mnt/_data_test_ type none (rw,bind)
> 
> Mount options are:
> Path: /data/test/

Can you try without the trailing slash?

> VFS type: loop
> Options: bind,rw
> 
> Cheers,
> Steve
> 
> On Tue, Aug 25, 2015 at 3:22 PM, Steve Kilduff < kild...@gmail.com > wrote:
> 
> 
> 
> Hi guys, trying to reply to a previous topic but I am certain I am not doing
> it correct, regardless, here we go.
> 
> moosefs mounting on ovirt.
> 
> I am having similar problem to the person in the mail Cannot add Posix
> Storage. This is an ovirt setup attempt with moosefs.
> 
> Mount options are:
> Path: mfsmount
> VFS type: fuse
> Mount opt: mfsmaster=mfsmaster,mfsport=9421,mfssubfolder=/ovirt,_netdev
> 
> I specified a subfolder which may not exist on your mfs storae.
> 
> In my setup, ovirt mounts the actual mfs mount without problem, the mount
> remains mounted afterwards as ovirt does not seem to unmount it
> successfully. I think the problem is the mount detection mechanism in
> python, It is looking for a pattern to match both the source/destination of
> the mount and failing: I hope this shows what I mean:
> vsdm.log output:
> OSError: [Errno 2] Mount of `mfsmount` at `/rhev/data-center/mnt/mfsmount`
> does not exist
> 
> centos7 output of mount command:
> mfsmaster:9421 on /rhev/data-center/mnt/mfsmount type fuse.mfs
> (rw,relatime,user_id=0,group_id=0,allow_other)
> 
> Source = mfsmount in first error out line, and source = ''mfsmaster:9421' on
> the OS, and I guess mount.py is trying to pattern match the mount and doesnt
> compare them correctly and bails out.

This seems to be the case.

ovirt needs a way to detect if a certain remote path is mounted at certain
mountpoint, so it compare the source of the mount and the mountpoint.
We don't have such issue with nfs or glusterfs, which do not change the name
of the remote path.

I would consult with the author of moosefs about this.

We can parse and ignore the ":port" suffix when comparing remote path, but
I'm not sure this if this behavior is common enough to add support for this.

For example, you can try to do this in the point where we compare the paths:

    if ":" in path:
        path = path.rsplit(":", 1)[0]

Before matching local path to remote path. If the path did not match before
because of the port suffix, it will much after removing the port suffix.

> 
> I guess normally the source of the mount does not change name, like with the
> specifics I gave to mount mfs.

Can you try different options to avoid the rename of the source?

For example, use the default port.

> 
> I am trying to poke around the python to see if I can simply ignore this
> check, but my python skills are non existent.
> 
> Also, as a workaround, I tried to first mount mfs without ovirt, and then get
> ovirt to bind mount my /mnt/moosefs how-ever in centos7 bind mount names are
> strange, I think broken when compared to centos6, the output of mount in
> centos7 when bind mounting /mnt/moosefs to /ovirt:
> mfsmaster:9421 on /mnt/moosfs
> mfsmaster:9421 on /ovirt
> 
> where as in centos6 it would be:
> mfsmaster:9421 on /mnt/moosfs
> /mnt/moosfs on /ovirt
> 
> Again, I think the pattern matching of mount.py is what fails during bind
> mount also.
> 
> I think if bind mounting worked, the user could stick what-ever filesystem
> they wanted onto the OS, and get ovirt to bind mount it. Then it's the
> user's problem how to manage their FS.
> 
> Best,
> Steve Kilduff
> 
> 
> 
> _______________________________________________
> Users mailing list
> Users@ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
> 
_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users

Reply via email to