Open Stack

On 07/14/2012 12:09 AM, Mark Moseley wrote:
> On Tue, Jun 12, 2012 at 8:20 AM, Pádraig Brady <P at draigbrady.com> wrote:
>> On 06/12/2012 04:07 PM, Fredric Morenius wrote:
>>> From: Scott Moser [mailto:ssmoser2 at gmail.com] On Behalf Of Scott Moser
>>> Sent: den 11 juni 2012 23:16
>>>
>>>> ...Without digging around on older versions of OS's and their included kernel/udev/kpartx.... i'd have to say that from your message above, and my testing that running 'kpartx' is no longer necessary.  That can reasonably be expected to be created by >the kernel or some other plumbing bits automatically...
>>>
>>> I just now tested to remove the kpartx invocation from /nova/virt/disk/mount.py, like this:
>>>
>>> diff --git a/nova/virt/disk/mount.py b/nova/virt/disk/mount.py
>>> index 11959b2..4d9527b 100644
>>> --- a/nova/virt/disk/mount.py
>>> +++ b/nova/virt/disk/mount.py
>>> @@ -61,25 +61,9 @@ class Mount(object):
>>>          if self.partition == -1:
>>>              self.error = _('partition search unsupported with %s') % self.mode
>>>          elif self.partition:
>>> -            map_path = '/dev/mapper/%sp%s' % (os.path.basename(self.device),
>>> -                                              self.partition)
>>> -            assert(not os.path.exists(map_path))
>>> -
>>> -            # Note kpartx can output warnings to stderr and succeed
>>> -            # Also it can output failures to stderr and "succeed"
>>> -            # So we just go on the existence of the mapped device
>>> -            _out, err = utils.trycmd('kpartx', '-a', self.device,
>>> -                                     run_as_root=True, discard_warnings=True)
>>> -
>>> -            # Note kpartx does nothing when presented with a raw image,
>>> -            # so given we only use it when we expect a partitioned image, fail
>>> -            if not os.path.exists(map_path):
>>> -                if not err:
>>> -                    err = _('partition %s not found') % self.partition
>>> -                self.error = _('Failed to map partitions: %s') % err
>>> -            else:
>>> -                self.mapped_device = map_path
>>> -                self.mapped = True
>>> +            #qemu-nbd already has mapped the partition for mounting
>>> +            self.mapped_device = '%sp%s' % (self.device, self.partition)
>>> +            self.mapped = True
>>
>> Looks good. That should also support raw through the /dev/loop%s devices.
>>
>> It might be safer to fallback to kpartx if not exists ...
>>   '%sp%s' % (self.device, self.partition)
>> That would support kernels before 3.2
>> I'd also add a comment in the code, that nbd must be mounted with param
>> max_part=16 or whatever.
> 
> 
> Might be totally unrelated to the above issues but I figured I'd throw
> it out there. I was messing with file injection today and was getting
> odd errors.
> 
> Some were due to hitting errors on a previous boot and partitions not
> getting cleaned up (e.g. kpartx -d never ran and the /dev/mapper/nbd*
> entries were still there, or qemu-nbd was still running). Some of
> those were likely a result of my own circumstances.
> 
> More interesting though, and what might be of use to other people, the
> "kpartx -a" calls get run and then the code in nova/virt/disk/mount.py
> immediately checks for whether or not the newly created
> /dev/mapper/nbdXXpX partitions exist. They'd actually get created but
> the os.exists call would fail. Apparently the os.exists call was
> getting run too soon. I added a time.sleep() after both the 'kpartx
> -a' and 'kpartx -d' calls, to give things time to catch up before the
> os.exists calls and things worked much better.

Sigh.

The amount of synchronization bugs I've noticed in lower
Linux layers lately is worrying.

Anyway I've created a bug for this:
https://bugs.launchpad.net/nova/+bug/1024586

cheers,
Pádraig