[DRBD-user] Effectiveness of barriers on disks with volatile write cache

Nicolae Mihalache xpromache at gmail.com
Fri Sep 24 11:12:35 CEST 2010


Hello,

I've been reading about the barriers (no-disk-barrier option) in drbd.
I understand that when the primary gets a IO completion notification,
it will issue a barrier request (actually start a new epoch) to the
secondary.
However, if the disk of the primary has a write cache, it will
immediately issue an IO completion notification, without actually
writing the data to the disk. So what happens is that the secondary
will use lots of barriers to guarantee the write order of the primary
while in fact the primary itself has no guarantee about the order.

My conclusion is in contradiction with what is written in the user
guide http://www.drbd.org/users-guide/re-drbdconf.html:

"When selecting the method you should not only base your decision on
the measurable performance. In case your backing storage device has a
volatile write cache (plain disks, RAID of plain disks) you should use
one of the first two (i.e. barrier or flush)."

Can someone point the fault in my reasoning?

thanks
nicolae


More information about the drbd-user mailing list