[DRBD-user] [SOLVED] suse 8.0 - drbd 0.6.12 - reiserfs kernel panic

Umberto Nicoletti unicoletti at prometeo.it
Wed Jun 23 17:09:07 CEST 2004

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

Hi all,
I post this message here only in the hope that it is helpful to somebody 
out there. If it is not your case please ignore it.

I am running a cluster on two IBM xSerier 235 servers, both running SuSE 
Linux 8.0 Professional, heartbeat 1.0.3, samba 2.2.8a, Pervasive SQL 8.1 
and drbd 0.6.12.

To setup drbd I followed the instructions found on this article:
linked on the linux-ha home page.

I use drbd to replicate the database files and samba shares (around 60GB 
on reiserfs 3.6.5). Heartbeat manages all these three processes plus a 
shared IP.

My problems began when I wanted to see what happened to the cluster if I 
pulled the power cord from the active node.
I was very disappointed to find out that on reboot the node would start 
giving kernel panics when it eventually tried to become master again and 
the database mounted the datafiles.
This happened repeatedly many times and on both nodes.

To make a long story short the culprit was in drbd configuration file.
In fact I made a (rather big ?) mistake when calculating the disk-size.

Following the directions on the before mentioned article I did the 

nodeb:~ # blockdev --getsize /dev/sda6
nodeb:~ # bc
bc 1.06
Copyright 1991-1994, 1997, 1998, 2000 Free Software Foundation, Inc.
This is free software with ABSOLUTELY NO WARRANTY.
For details type `warranty'.
131552379 / 2

so I set disk-size to 65776189.

In fact that was wrong, because if I try to format the partition with 
ext3 it complains about the fact that it cannot mark the end of the 
device, which reiserfs doesn't. At least until it needs to replay logs 
on a uncleanly unmounted filesystem, at which points it panics out.

So I set disk-size to 65776188 and that solved all my problems and now 
the cluster is surviving happily to all kind of nasty events such as 
power failures or network erros.

To conclude I would like to send my heartfelt thanks to the people 
behind drbd and heartbeat.

Thanks again for you attention,

More information about the drbd-user mailing list