Description of problem: The gulm initscript should run only if a gulm configuration is detected. The script will need to check /etc/cluster/cluster.conf to determine if gulm is configured since ccsd will not be in a quorate state when this script is started on bootup. The check should silently cause the initscript to exit so that users are not alarmed by unneccessary error messages. Version-Release number of selected component (if applicable): problem exits in gulm-1.0-0.pre11.0 and earlier.
diff -u -p -r1.1.2.4 lock_gulmd --- cluster-RHEL4/gulm/init.d/lock_gulmd 12 Jan 2005 16:00:26 -0000 1.1.2.4 +++ cluster-RHEL4/gulm/init.d/lock_gulmd 20 Jan 2005 17:55:34 -0000 @@ -36,6 +36,21 @@ then GULM_OPTS="$GULM_OPTS --use_ccs" fi + +# If gulm is using ccs, then make sure that there is a gulm +# section in /etc/cluster/cluster.conf, otherwise abort. +# FIXME -- Should this be silent? I think users should get some +# feedback, but others might not want added verbosity to +# the boot process. Oh well... it's only one line :) +if echo "$GULM_OPTS" | grep -qE "(--use_ccs|[\t ]-[VhCed]*c)" +then + if ! grep -qE "<[\t ]*gulm[\t ]*.*>" /etc/cluster/cluster.conf + then + echo "lock_gulmd: no <gulm> section detected in /etc/cluster/cluster.conf" + exit 11 + fi +fi + /cvs/cluster/cluster/gulm/init.d/lock_gulmd,v <-- lock_gulmd new revision: 1.5; previous revision: 1.4 /cvs/cluster/cluster/gulm/init.d/lock_gulmd,v <-- lock_gulmd new revision: 1.1.2.5; previous revision: 1.1.2.4
Verified in gulm-1.0-0.pre16.0. Starting lock_gulmd: no <gulm> section detected in etc/cluster/cluster.conf [FAILED]