[DRBD-user] drbd0: [kjournald/1476] sock_sendmsg time expired, ko = 4294967295

Lars Ellenberg Lars.Ellenberg at linbit.com
Wed Oct 25 14:56:51 CEST 2006

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


/ 2006-10-25 14:28:46 +0200
\ Leroy van Logchem:
> 
> >/ 2006-10-21 14:14:34 +0200
> >\ Sim:
> >  
> >>Hi!
> >>
> >>Today I have see this and the server was been locked:
> >>
> >>Oct 21 11:28:51 mx kernel: drbd0: [kjournald/1476] sock_sendmsg time
> >>expired, ko = 4294967295
> >>    
> >...
> >  
> >>Oct 21 11:29:50 mx kernel: drbd0: [kjournald/1476] sock_sendmsg time
> >>expired, ko = 4294967283
> >>    
> >
> >google for drbd ko-count ...
> >I assume your Secondaries IO-subsystem got stuck.
> >
> >you should configure ko-count = 6 or so.
> >it would have started to count ".... ko = 6",
> >and at "ko = 0" it would have gone StandAlone,
> >recovering from this situation.
> >
> >you should also upgrade to a 2.6 kernel,
> >where this won't lock up the box but only the drbd partition.
> >2.4 has only _one_ runqueue for _all_ disks, if it gets stuck
> >on one disk, all io is stuck.
> >  
> Same here. Running a 2.6 kernel without ko-count. We see about 400
> sendmsg_time expired per day, is there a tuning method to reduce
> these?

get network equipment that can handle your traffic, or alternatively
disks that can cope with what your network delivers :)
maybe using a different io-scheduler helps, too (I still prefer deadline).

otherwise, the timeout parameter in drbd.conf is your tunable.
note that its unit is 0.1 seconds, the default is 60 (== 6 seconds).

maybe using larger (or smaller) sndbuf-size makes a difference, too.

-- 
: 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