CentOS 5.5 x86_64 (2.6.18-194.32.1.el5), drbd 8.3.8-1.
I am concerned about the usefulness and/or accuracy of the drbd verify
operation. Running 'drbdadm verify' on a resource, I get these messages in
the log:
Feb 20 20:02:29 tiger kernel: block drbd11: Out of sync: start=622335312,
size=8 (sectors)
Feb 20 20:02:34 tiger kernel: block drbd11: Out of sync: start=623124784,
size=8 (sectors)
...
Feb 21 02:04:48 tiger kernel: block drbd11: Online verify done (total 25857
sec; paused 0 sec; 75504 K/sec)
Feb 21 02:04:48 tiger kernel: block drbd11: Online verify found 6 4k block out
of sync!
Feb 21 02:04:48 tiger kernel: block drbd11: conn( VerifyS -> Connected )
Feb 21 02:04:48 tiger kernel: block drbd11: Writing the whole bitmap, due to
failed kmalloc
Feb 21 02:04:49 tiger kernel: block drbd11: 24 KB (6 bits) marked out-of-sync
by on disk bit-map.
Observations and problems:
- It says "Online verify found 6 4k block out of sync" whereas there were
actually 25 "Out of sync:" messages.
- The verify was run after a re-sync of the whole device, so presumably
there should be no positives, real or false, at all.
- If I repeat the verify, it will show a different number of out of sync
blocks at different places. Same for a further repeat, etc.
Given this, and also that I have seen several other people report a
similar situation, it appears impossible to use the verify operation in
real life, and further it appears impossible to find out if a resource is
really in sync. I would hope that the drbd developers and other
knowledgeable folk would comment on this.
Steve
_______________________________________________
drbd-user mailing list
[email protected]
http://lists.linbit.com/mailman/listinfo/drbd-user
