HP Serviceguard Storage Management Suite Version A.02.00 for HP-UX 11i v2 Release Notes, September 2008

JAGag17495: fsclustadm cfsdeinit device busy, package halt timeout
• What is the problem? In some stress situations, during cmhaltcl, the package halt reaches
time out. This causes the node to perform a transfer of control (TOC). This problem is caused
by an error in the CFS shutdown process.
• Workaround? There is no workaround. The failing node should recover after the TOC.
JAGag20605: cvm join problem
• What is the problem? In some stress situations, when a node starts up, the Serviceguard System
Multi-node package fails to start due to a CVM join failure. This causes the joining node to
perform a TOC.
• What is the workaround? There is no workaround. When the node recovers from the TOC,
run the cmrunnode CFS command. It should complete successfully.
JAGag22707: Lot's of vx* directories left over in /tmp
• What is the problem? Some empty directories can be seen in /tmp/vx* after a reboot.
• What is the workaround? These directories can be ignored.
JAGag35297: cfsmount failed since the package was already started by cmcld
• What is the problem? This affects the CFS mount point package. If the package is already
down on all nodes when you run the cfsmount command, cfsmount could fail, complaining
that the package has already been started.
• What is the workaround? The workaround is to issue cfsmntadm display <mount_point>
to verify if the mount point was mounted successfully and re-issue cfsmount command if
retry is needed.
1. After running cfsmount, use cfsadmin to verify that the mount succeeded:
cfsmntadm display <mount_point>
2. If the mount failed, re-run cfsmount.
JAGag37114: Diskgroup resync taking too long
• What is the problem? With multiple disk groups in the cluster, it is possible that mirrored
volumes in more than 1 diskgroup require syncing. When the cluster comes up, CVM initiates
the vxrecover on the master node. However when the node is joining the cluster,
ServiceGuard script also checks the kernel state of the volumes in the shared disk groups.
If the kernel state of any volume in the shared diskgroup is DISABLED, then Serviceguard
calls a vxvol start on the master node. This can happen if vxrecover still has not started
the recovery of that diskgroup. The vxvol start command does the syncing of volumes
one at a time and in the foreground. If many volumes need syncing and if the system is
heavily loaded, then this syncing can take a long time.
• What is the workaround? There is no workaround.
JAGag40401: cannot sleep during interrupt
• What is the problem? In some stress situation where the system is under heavy IO load, if
some nodes fail (TOCs), CVM resyncs the mirrored shared volumes for the surviving nodes.
While CVM is resync'ing the volumes, it might invoke sleep(). If this sleep() is invoked
during an interrupt context, the system performs a TOC.
• What is the workaround? There is no workaround.
28 HP Serviceguard Storage Management Suite Version A.02.00 Release Notes