HP Serviceguard for Linux Version A.11.16 Release Notes, Third Edition, August 2006

Serviceguard for Linux Version A.11.16 Release Notes
Patches and Fixes in this Version
Chapter 1 51
JAGaf01170: cmresmond should kill agents during
shutdown/restart/reload
Problem: On cmresmond --shutdown, --restart, or --reload, the disk
agents that were running before the command was executed may not be
terminated and will be left running. The disk agents are never explicitly
killed in these cases because they usually die on their own.
Resolution: Kill agents explicitly upon reload, restart, or shutdown.
JAGaf09870: cmresmond core - something died in cmresmond
somewhere
Problem: When cmresmond is running, and "cmresmond --reload" is
executed, cmresmond will terminate with a segmentation fault. SIGCHLD
handling does not behave correctly during reload. The signal handler is
closed without ignoring signals, so when the SIGCHLD occurs during the
reload window, the segmentation fault occurs.
Resolution: Ignore signals before closing signal handler.
JAGaf11607: cmcld on linux will abort if the deadman driver is
not loaded
Problem: cmcld will abort if the deadman driver is not loaded and the
cluster is started. There is a small (and unlikely) window that occurs at
system boot up time that may result in a node not being able to join the
cluster. This happens if a cmrunnode from another node is issued just
after the node rebooting starts its network services and before the
deadman driver is loaded.
Resolution: Load the deadman driver before networking is enabled on
system boot up. Divide cmcluster.init into two files. The first is called
SGSafety. This loads the deadman driver before any network services
are started at boot up.
JAGaf11704: cmmodnet goes into infinite loop when zero subnet
specified
Problem: cmmodnet gets stuck when a zero subnet (instead of a valid
subnet) is passed as an argument to it which would thus cause the
package start to hang when issued from the package control script.