[DRBD-user] DRBD stalls reproducibly on every "drbdadm verify"

Lars Ellenberg lars.ellenberg at linbit.com
Thu Jan 9 16:46:03 CET 2014

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

On Thu, Jan 09, 2014 at 10:14:04AM +0100, Christian Hammers wrote:
> Hello
> I found the following knowledge base article from 2011 which described my
> exact problem: http://www.novell.com/support/kb/doc.php?id=7009306
> The solution given there was to switch from a static to an adaptive sync
> rate, especially when using very fast (gigabit) network interface cards.
> For me, it seemed to work when switching from "rate 40M" to:
>   syncer {
>         c-plan-ahead 20;
>         c-min-rate 1M;
>         c-max-rate 300M;
>         c-fill-target 2M;
>         verify-alg md5;
>   }
> Before that I also tried to use a plain crossover interface instead of the
> bonding one but that had no effect. Adjusting the other values recommended
> in the KB article did work for me, too, but I changed them back to their
> defaults to isolate the above mentioned as the real fix.
> Any comments on this one? Is this a bug in DRBD?

The root cause is a distributed resource starvation deadlock on our
"max-buffers" setting.

The real fix is to upgrade to >= 8.4.4.

The previous workaround is to not exhaust max-buffers with
Whether you achieve that by increasing max-buffers,
or by starting the verify from the other node, or by reducing the number
of queued resync requests per unit time, or both, is not that important.

: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
please don't Cc me, but send to list   --   I'm subscribed

More information about the drbd-user mailing list