[linux-lvm] mirrored LV + cmirror problem
Jonathan Brassow
jbrassow at redhat.com
Mon Feb 18 18:30:49 UTC 2008
That's the problem. (C)LVM does not currently have a way to handle
the disappearance of a device from just one machine... it expects a
similar view of devices from all machines in a cluster. "Locking" on
the second node fails because it doesn't know what to do with the disk
that it sees (that has been removed from the node suffering the
failure).
There is someone working handling orphaned/removed devices that
reappear (which would be similar to this case), but I'm not sure if
they've taken this scenario into account. I'll let him know about this.
brassow
On Feb 18, 2008, at 10:48 AM, Lajkó Attila wrote:
> Yes, I disconnected the lun via the Vtrak GUI from one of the nodes
> (el4).
>
> Attila
>
> On Feb 18, 2008, at 5:08 PM, Jonathan Brassow wrote:
>
>> How are you performing the failures? It looks like just one
>> machine is loosing its connection to the device, while the other
>> machines links remain in place.
>>
>> brassow
>>
>> On Feb 17, 2008, at 6:49 AM, Lajkó Attila wrote:
>>
>>> Here is the output of vgreduce:
>>> ftp://ftp.ulx.hu/upload/clvmd/vgreduce.tar.gz
>>>
>>> Attila
>>>
>>>
>>> 2008. febr. 15, 22:06 DU dátummal Jonathan Brassow <jbrassow at redhat.com
>>> >
>>> ezt írta:
>>>
>>>> If the problem is reproducible, we should be able to track it down.
>>>>
>>>> When a failure happens, the kernel sends an event to userspace that
>>>> signals 'dmeventd' to take action. If we take dmeventd out of the
>>>> picture, we can run the commands ourselves with higher verbose
>>>> settings.
>>>>
>>>> When you activate the volume, you can 'lvchange -ay --monitor n
>>>> <vg>/
>>>> <lv>' - this will prevent dmeventd from monitoring the mirror. Then
>>>> kill the log device. Finally, run 'vgreduce --removemissing <VG> -
>>>> vvvv' to perform the recovery. (redirecting all the output to a
>>>> file
>>>> will give us something to look at if the failure is reproduced.)
>>>>
>>>> We may need to grab debugging output from clvmd too, but that can
>>>> get
>>>> messy, so we'll start with this.
>>>>
>>>> brassow
>>>>
>>>> P.S. It looks like you must have *.debug; in your /etc/syslog.conf,
>>>> yes?
>>>>
>>>> On Feb 15, 2008, at 10:17 AM, Lajkó Attila wrote:
>>>
>>>
>>>
>>> _______________________________________________
>>> linux-lvm mailing list
>>> linux-lvm at redhat.com
>>> https://www.redhat.com/mailman/listinfo/linux-lvm
>>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>>
>>
>> _______________________________________________
>> linux-lvm mailing list
>> linux-lvm at redhat.com
>> https://www.redhat.com/mailman/listinfo/linux-lvm
>> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
>
>
> _______________________________________________
> linux-lvm mailing list
> linux-lvm at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-lvm
> read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/
More information about the linux-lvm
mailing list