HP Serviceguard for Linux Version A.11.18 Release Notes, 2nd Edition, March 2009

program cmclconfd is using a deprecated SCSI ioctl, please
convert it to SG_IO
cmclconfd uses SCSI ioctl calls to the lock LUN device to get various pieces of
information. The ioctl currently being used is being deprecated but still works
correctly.
What is the workaround? (This is an informational message only and can safely be
ignored.)
JAGag44319 (SR8606492148): cmcluster boot script does not print cmrunnode errors
• What is the problem? If AUTOSTART_CMCLD is set to 1 and cmrunnode starts the
cluster successfully after a node reboot, connectivity errors on startup, if any, are
not logged to /etc/rc.log.
• What is the workaround? The cmrunnode errors (but not the command
log) can be found in the syslog file.
QXCR1000740150: cmcld abort when deleting node online
• What is the problem? When you run cmapplyconf to delete a node from the cluster
configuration while the cluster is running, cmcld aborts if the name of the node
cmapplyconf is run from is a subset of the name of the node being deleted.
For example, if you run cmapplyconf from node1 to delete node10, cmcld
aborts with the following message in syslog.log:
cmcld: Aborting! Deleting local node that is online
(file: rcomm/comm_config.c, line: 583)
The cluster daemon aborted our connection (231).
cmapplyconf: Unable to apply the configuration
• What is the workaround? Avoid the problem by naming nodes such that no node’s
name is a subset of another’s.
QXCR1000741571: Hanging node problem causing cluster to go down
• What is the problem? A hang on one node can cause multiple cluster re-formations,
until eventually the entire cluster fails.
• What is the workaround? There is no workaround at present.
QXCR1000749924: Dependent packages started on non-primary nodes
• What is the problem? Failover packages which are dependent on a multi-node
package may start on a node other than their primary node even though the
dependency is met on the primary node.
• What is the workaround? One workaround for this problem is to set the dependent
packages’ failback_policy to automatic, so if the package starts on a non-primary
node, it will be automatically failed back as soon as the primary node is available.
84 Serviceguard for Linux Version A.11.18 Release Notes