Serviceguard for Linux support of the DL585 and cmresmond

[Serviceguard for Linux]

During Serviceguard for Linux certification testing of the DL585 with RedHat EL 3 U2, a seldom seen problem was discovered. This only affects the disk monitor (cmresmond) and is not specific to the DL585. If this problem is encountered cmresmond will exit with a 'core dump'. This would cause a failover of any package that uses the monitor.

While a patch to address this issue will be available soon, we wanted to allow customers to build Serviceguard clusters with the DL585 immediately. A workaround for this issue is to install and activate a new version of cmresmond.

1. To activate the new version of cmresmond make sure there are no packages running on the node that use it (either halt the packages or move them to another node).

2. Check the PID of cmresmond
#cmresmond --status

3. Relocate the existing cmresmond:
# cd /usr/local/cmcluster/bin
# mv cmresmond cmresmond.orig

4. Download the attached file named cmresmond put it in /usr/local/cmcluster/bin

5. Check the build date of the new cmresmond by running the command
# strings cmresmond | grep 'Build date'
And see that the build date is July 30, 2004

6. Restart cmresmond:
# cmresmond --restart

7. Check that the PID of cmresmond is different than displayed in step #2
#cmresmond --status

8. Restart (or relocate) any packages.

9. Repeat steps 1-8 for all nodes in the cluster.

Note: the options shown for cmresmond (status and restart) have a space and two dashes before them. It may not be clear on the screen.

Regards,
The Serviceguard for Linux Team

[Serviceguard for Linux]

The DL585 is now supported with SUSE SLES8 (and UnitedLinux 1.0). The configuration guide (ftp://ftp.compaq.com/pub/solutions/enterprise/ha/linux/svcguard-certmatrix.pdf) will have the supported errata details. There is a DIFFERENT version of cmresmond for SUSE/UL. It is attached to THIS part of the posting. The instructions are the same as above except for the directory noted in steps 3 and 4. In this case the directory is /opt/cmcluster/bin

Regards,
The Serviceguard for Linux Team