[DRBD-user] What happen if the Master(or Primary) server disk crash during read/write file on client request.

Seki Lau sekilau at gmail.com
Wed Sep 22 22:18:40 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.


Dear All,
I worked for a week to set up a 640G HA file server. The application
for accept users read/write is samba. When failover, secondary(now
becomes primary) mounts the drbd disks to mount points and starts the
smbd,nmbd. The smb.conf on both machines are same. So ordinary clients
will not notice the failover(unless they are accessing files during
failover and it is the primary machine fail).

I notice there has tool CTH comes with the drbd0.7.4, but because I
want to understand those scripts before I use them and the deadline
for the fileserve is coming I decide to draft test cases myself.


I would like to simulate the case that, the heartbeat between primary
and secondary is still exist but it is only the disk in primary crash
(bad sector or even read/write head fail). Therefore , I simulate it
as follow :
On the primary machine(with drbd0,drbd1 mounted), during copy a file
from drbd0 to drbd1, kill the job related to drbd1 or drbd0(or even
both drbd0 and drbd1).

The person I work for accepts recopy the file if the primary server
fail during the file access.

But I wonder will the primary machine fail and become secondary if
drbd0/drbd1 related job is kill in the middle of file read/write.

Can any one give me some idea?
Does on-error detach or on-error panic in drbd.conf can achieve this?

Thanks in prior
Seki Lau



More information about the drbd-user mailing list