[DRBD-user] one of drbd disk need full resync after reboot

chatchai jantaraprim chatchai.j at gmail.com
Wed Jun 3 07:51:22 CEST 2009

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Tue, Jun 2, 2009 at 9:15 PM, Lars Ellenberg
<lars.ellenberg at linbit.com> wrote:

> may or may not be fixed in 8.0.16.
> may or may not be fixed in 8.0.git
> may or may not be fixed in 8.3(.git).
>
> as you apparently are able to reproduce this,
> could you verify the above?

  Umm, these servers are production servers which 200+ users rely on them.
It is difficult to get schedule downtime for maintainance. It will
really difficult
to get experiment on these servers.
  However what we are using on these machines are stock kernel and modules
from debian lenny distribution. So, it might possible to setup two PC with the
same kernel and drbd module, and try to change different drbd modules.
  However I'm not sure that the number of drbd devices or the size of drbd
devices may have effects on the event of full resync or not. On the production
servers we have 7 drbd devices and each 80GB+ in size. We didn't have
spare disks of that size to test. The fully resync occurred only on the last
drbd disk, the other drbd devices didn't.

> if you can still reproduce,
> would you be so kind and give the minimum procedure to reproduce?

  If we can setup the test machines, and reproduce this, I'll send you
more result.

> try to avoid line breaks when pasting log files.
> they are horrilbe to read this way.

   Sorry for that, I forgot the recheck the result before sending.
I use gmail, is it okay if I send it using "Rich formatting" to
make sure that lines is not wrapped?

> this may be a race condition in drbd.
> we will double-check.

  We are using the older kernel version for sometimes, I'm not sure
about version of drbd, without seeing this. Well, may be, may be not,
but it seems that those when it need full resync, all of drbd devices
will need at the same time after reboot, so that might our operational
problem. But these times, only single drbd device that has this full resync.

>
>> uuid_compare()=0 by rule 4
>> May 28 13:21:09 nnksvr01 kernel: [  150.504885] drbd6: No resync, but
>> 44522728 bits in bitmap!
>
> there.
> why do you have that many bits (in fact: all bits) set?
>
>>

Cheers,
cj



More information about the drbd-user mailing list