[DRBD-user] drbd network block device deadlock ?

Philipp Reisner philipp.reisner at linbit.com
Thu Feb 3 10:29:39 CET 2011

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


Am Dienstag, 1. Februar 2011, um 16:49:05 schrieb Ivan Frain:
> Hi all,
> 
> I am currently evaluating DRBD as a storage candidate for highly
> available storage in a virtualized environment.
> It seems like a very good alternative to expensive SAN/NAS.
> I was wondering how DRBD deals with the network block device deadlock
> problem.
> This problem (described here: http://lwn.net/Articles/195416/) can be
> summarized as follows: if the system runs short in memory, it will try
> to write dirty page to disk in order to free memory space.  if the
> disk is a network block device, the dirty page write may need to
> allocate some other memory pages which is not possible since the
> solution to have more memory available was to write the dirty page to
> disk.
> 

This is a very well understood issue. The solution to the issue is, that
all memory allocations in the write out path are backed by mempools
(= Small amounts of pre allocated memory).
DRBD does not deadlock in out of memory situations, at least since version
8. Maybe even earlier.

Best,
 Phil
-- 
: Dipl-Ing Philipp Reisner
: LINBIT | Your Way to High Availability
: Tel: +43-1-8178292-50, Fax: +43-1-8178292-82
: http://www.linbit.com

DRBD(R) and LINBIT(R) are registered trademarks of LINBIT, Austria.



More information about the drbd-user mailing list