Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hi,
Has anyone successfully run cyrus-imapd on DRBD partitions?
In my case, I have set up 2 identical nodes running Suse 9
(2.6.5-7 kernel) using the bundled drbd-0.7.0-59.22 and
heartbeat-1.2.2-0.6.
Copying and deleting several hundred MBs repeatedly on the DRBD
partitions (reiserfs) worked fine.
But, as soon as I configured cyrus-imapd "configdirectory" and
"partition-default" to use the DRBD partitions, I got these messages:
master[4506]: process started
master[4507]: about to exec /usr/lib/cyrus/bin/ctl_cyrusdb
kernel: drbd0: Local IO failed. Detaching...
kernel: drbd0: Local IO failed. Detaching...
kernel: drbd0: Local IO failed. Detaching...
kernel: drbd0: Local IO failed. Detaching...
kernel: drbd0: local read failed, retrying remotely
kernel: drbd0: local read failed, retrying remotely
kernel: drbd0: local read failed, retrying remotely
kernel: drbd0: local read failed, retrying remotely
kernel: drbd0:
/usr/src/packages/BUILD/kernel-smp-2.6.5/modules-2.6.5/drbd/drbd_actlog.
c:649: Connected flags=0x5509
kernel: drbd0:
/usr/src/packages/BUILD/kernel-smp-2.6.5/modules-2.6.5/drbd/drbd_actlog.
c:649: Connected flags=0x5509
[this last messages repeat non-stop]
The slave node shows the partition as consistent but
the primary node /proc/drbd showed:
version: 0.7-pre8 (api:74/proto:72)
0: cs:DiskLessClient st:Primary/Secondary ld:Inconsistent
ns:0 nr:0 dw:684600 dr:2008 al:360 bm:0 lo:0 pe:0 ua:0 ap:0
DRBD never recovered from this state and the system would
eventually hang. When I mount the SAN-disks directly without
using DRBD, cyrus-imapd worked fine.
Any idea how I could go about troubleshooting this?
Details of my setup:
- Suse Enterprise Linux 9 with 2.6.5-7.111-smp kernel
- DRBD partitions mounted on fiber-channel connected SAN disk
(IBM shark) using QLogic FC HBA.
- Each node has two ethernet connections (switched, not x/over),
100Mbps and 1Gbps. 1Gbps is used for heartbeat and DRBD.