[DRBD-user] Congratulations and problems...

[Lars Ellenberg]

On Mon, Feb 16, 2004 at 11:24:01AM +0100, Felix Ide wrote:
> Hi Lars,
> 
> I have increased ping-int to a really lousy large value (just for the fun of 
> it ;) and now the primary system stalled, no more entries in syslog, 
> nothing... :-(

> The last entry in syslog was: 
> Feb 15 21:18:34 drbd1-bfd kernel: drbd0: [kupdated/7] sock_sendmsg returned 
> -32

again, what does the syslog on the peer say?

> Feb 15 21:18:50 drbd1-bfd kernel: drbd0: Connection lost.
> Feb 15 21:18:50 drbd1-bfd kernel: drbd0: Connection established. size=26097472 
> KB / blksize=4096 B
> Feb 15 21:18:50 drbd1-bfd kernel: drbd0: Synchronisation started blks=32
> Feb 15 21:18:53 drbd1-bfd kernel: drbd0: transferlog too small!! 

oops!
increase it!

> Feb 15 21:18:55 drbd1-bfd kernel: drbd0: [kupdated/7] sock_sendmsg returned 
> -104
> Feb 15 21:19:23 drbd1-bfd kernel: drbd0: Syncer aborted.
> Feb 15 21:19:28 drbd1-bfd kernel: drbd0: Connection lost.
> Feb 15 21:19:29 drbd1-bfd kernel: drbd0: Connection established. size=26097472 
> KB / blksize=4096 B
> Feb 15 21:19:29 drbd1-bfd kernel: drbd0: Synchronisation started blks=32
> Feb 15 21:30:24 drbd1-bfd kernel: drbd0: Synchronisation done.
> 
> the next -- MARK -- message should have been written at 21:43, but it wasn't. 
> This time 5 bonnie++ processes were running.
> 
> drbd1-bfd:~ # drbdsetup /dev/nb0 show
> Lower device: 09:01   (/dev/md1)
> Disk options:
>  do-panic
> Local address: 10.20.30.1:7788
> Remote address: 10.20.30.2:7788
> Wire protocol: C
> Net options:
>  timeout = 600.0 sec
 
 ok, go back to default values here, 6.0 or smaller.
 
>  ko-count = 20

did you ever see some "ko-count=19 ... 18 ... 17" messages?

>  tl-size = 10000     <== still too small :(


>  connect-int = 601 sec
>  ping-int = 601 sec
>  sndbuf-size = 262144    <== increase it.

it is "safe" to increase it with protocol C.

>  sync-min = 500 KB/sec
>  sync-max = 204800 KB/sec
> 
> Thanks,
> Felix