[DRBD-user] What indication of outstanding writes when WFConnection?

Lars Ellenberg Lars.Ellenberg at linbit.com
Tue Jul 26 16:11:00 CEST 2005

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


/ 2005-07-26 14:51:38 +0200
\ Leroy van Logchem:
> Mark Yarborough wrote:
> 
> >Hello all,
> >Can someone clarify the warnings you might get when running protocol
> >C and a write to the primary cannot be replicated to the secondary?
> >Like so:
> >1. Primary and Secondary both cs:Connected
> >2. Disconnect DRBD line (only)
> >3. Primary and Secondary both cs:WFConnection
> >4. Write to DRDB partition on primary
> >5. ???????
> >6. Reconnect DRBD line

you forgot the resync
 (cs:WFBitmapS, PausedSyncS, SyncSource etc.)
fortunately in the described scenario, the resync typically is that fast
that you should not see these states for too long in /proc/drbd.
do grep -i "drbd.*sync" /var/log/kern.log 
or similar...

> >7. Primary and Secondary both cs:Connected
> >At step 5 the primary and secondary partitions were out of sync, because the DRBD line was down.
> >From the primary, at step 5, is there any way to tell
> >that the partitions are out of sync? Any message logged?

what kind of message do you want to be logged?

> IMO: The drbd state machine always marks the device out-of-sync if
> disconnected from the primary ( even if it's only for a short while ).
> Your primary mounted filesystem will always change a bit if it where
> only the timestamps. From WFConnection to Connected the state is
> exchanged and typically the next state is resync. So just monitoring
> for anything other then "Connected" when issued a "drbdadm cstate all"
> should be enough to report a event using Nagios/NetSaint.

there is /proc/drbd, too.
additionally, on cstate changes there is printk (kernel log) output like
kernel: drbd0: drbd0_worker [18590]: cstate SyncSource --> Connected
        drbd<minor> <cmd> [<pid>]: cstate <FromState> --> <ToState>
so a simple filter and sync(0) definition in syslog-ng and a proper
logcheck rule should do.
that is much better than polling with drbdadm cstate, I think.

-- 
: Lars Ellenberg                                  Tel +43-1-8178292-0  :
: LINBIT Information Technologies GmbH            Fax +43-1-8178292-82 :
: Schoenbrunner Str. 244, A-1120 Vienna/Europe   http://www.linbit.com :
__
please use the "List-Reply" function of your email client.



More information about the drbd-user mailing list