[Users] Error creating the first storage domain (NFS)

[Brian Vetter]

Ugh. Spoke a little too soon. While I got past my problem creating a storage domain, I ran into a new sanlock issue.

When trying to run a VM (the first one so I can create a template), I get an error in the admin UI:
VM DCC4.0 is down. Exit message: Failed to acquire lock: Permission denied.

The sanlock.log file shows the following:
2012-10-23 22:32:02-0500 22023 [981]: s3:r3 resource 8798edc0-dbd2-466d-8be9-1997f63e196f:71252c8f-68a9-495f-b5a6-4e8e035b56ea:/rhev/data-center/a8ea368c-bc08-4e10-81e7-c8439bf7bd35/8798edc0-dbd2-466d-8be9-1997f63e196f/images/b029b5a6-9eb3-4a34-ad03-1ac4386e8c7c/71252c8f-68a9-495f-b5a6-4e8e035b56ea.lease:0 for 2,11,14629
2012-10-23 22:32:02-0500 22023 [981]: open error -13 /rhev/data-center/a8ea368c-bc08-4e10-81e7-c8439bf7bd35/8798edc0-dbd2-466d-8be9-1997f63e196f/images/b029b5a6-9eb3-4a34-ad03-1ac4386e8c7c/71252c8f-68a9-495f-b5a6-4e8e035b56ea.lease
2012-10-23 22:32:02-0500 22023 [981]: r3 acquire_token open error -13
2012-10-23 22:32:02-0500 22023 [981]: r3 cmd_acquire 2,11,14629 acquire_token -13

I looked at the lease file referenced above, and it is there.
[root at mech ~]# ls -l /rhev/data-center/a8ea368c-bc08-4e10-81e7-c8439bf7bd35/8798edc0-dbd2-466d-8be9-1997f63e196f/images/b029b5a6-9eb3-4a34-ad03-1ac4386e8c7c/71252c8f-68a9-495f-b5a6-4e8e035b56ea.lease
-rw-rw----. 1 vdsm kvm 1048576 Oct 23 22:30 /rhev/data-center/a8ea368c-bc08-4e10-81e7-c8439bf7bd35/8798edc0-dbd2-466d-8be9-1997f63e196f/images/b029b5a6-9eb3-4a34-ad03-1ac4386e8c7c/71252c8f-68a9-495f-b5a6-4e8e035b56ea.lease

On a lark, I turned off selinux enforcement and tried it again. It worked just fine.

So what selinux option do I need to enable to get it to work? The only other sanlock specific settings I saw are:
sanlock_use_fusefs --> off
sanlock_use_nfs --> on
sanlock_use_samba --> off

Do I turn these all on or is there some other setting I need to enable?

Brian

On Oct 23, 2012, at 9:54 PM, Brian Vetter wrote:

> That was the problem. I checked the sanlock_use_nfs boolean and it was off. I set it and then created and attached the storage and it all works.
> 
> Thanks for the help/pointer.
> 
> Brian
> 
> On Oct 23, 2012, at 8:55 PM, Federico Simoncelli wrote:
> 
>> Hi Brian,
>> I hate progressing by guesses but could you try to disable selinux:
>> 
>> # setenforce 0
>> 
>> If that works you could go on, re-enable it and try something more
>> specific:
>> 
>> # setenforce 1
>> # setsebool sanlock_use_nfs on
>> 
>> I have the feeling that the vdsm patch setting the sanlock_use_nfs
>> sebool flag didn't made it to fedora 17 yet.
>> -- 
>> Federico
>> 
>> ----- Original Message -----
>>> From: "Brian Vetter" <bjvetter at gmail.com>
>>> To: "Federico Simoncelli" <fsimonce at redhat.com>
>>> Cc: "Vered Volansky" <vered at redhat.com>, users at ovirt.org, "David Teigland" <teigland at redhat.com>
>>> Sent: Tuesday, October 23, 2012 6:10:36 PM
>>> Subject: Re: [Users] Error creating the first storage domain (NFS)
>>> 
>>> Ok. Here's four log files:
>>> 
>>> engine.log from my ovirt engine server.
>>> vdsm.log from my host
>>> sanlock.log from my host
>>> messages from my host
>>> 
>>> The errors occur around the 20:17:57 time frame. You might see other
>>> errors from either previous attempts or for the time after when I
>>> tried to attach the storage domain. It looks like everything starts
>>> with an error -13 in sanlock. If the -13 maps to 13/EPERM in
>>> errno.h, then it is likely be some kind of permission or other
>>> access error. I saw things that were related to the nfs directories
>>> not being owned by vdsm:kvm, but that is not the case here.
>>> 
>>> I did see a note online about some issues with sanlock and F17 (which
>>> I am running), but those bugs were related to sanlock crashing.
>>> 
>>> Brian
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20121023/963d3e0d/attachment-0001.html>