HP Serviceguard Version A.11.17 Release Notes, March 2006 (revised)

Serviceguard Version A.11.17 Release Notes
Known Problems and Workarounds
Chapter 1 53
What is the workaround? If the overall disk configuration (number of
LUNs) is not large, wait for the command to complete, otherwise
terminate the command with CTRL-C.
JAGaf36760 (SR8606376483): Uncorrectable write
errors on vxvm volumes.
What is the problem? Configurations using greater than 256 LUNs
with VxVM on 11i v2 September 2004 update experience
uncorrectable write errors.
What is the workaround? Restrict configurations to a maximum of
256 LUNs. Please continue to check the status of this problem at
http://www.itrc.hp.com for updated information. Search the Technical
Knowledge base for bug reports with the keyword JAGaf36760.
JAGaf32484 (SR8606372066): Problems doing a rolling
upgrade when moving from CVM 3.2 to 3.5
What is the problem? In a Serviceguard cluster with CVM 3.2, it is
not possible to perform a rolling upgrade from CVM 3.2 to CVM 3.5.
After a node is upgraded to CVM 3.5, when it attempts to rejoin the
CVM cluster, it will fail, possibly causing the node to be TOC'ed.
This problem exists regardless of whether the HP-UX operating
system revision was changed during the rolling upgrade. By "rolling
upgrade", we mean halting one node in the cluster, upgrading CVM
to 3.5 and then starting the node back in the cluster, then performing
the same steps for each node in the cluster. The cluster is never
completely halted. However, this issue does require the cluster to be
completely halted for at least some period of time, since it is not
possible to run the cluster with a mix of CVM 3.2 and 3.5.
What is the workaround?
A workaround exists to do a semi-rolling upgrade to a cluster being
upgraded from CVM 3.2 to 3.5. It requires a small amount of
downtime though.
This workaround involves managing the nodes in the cluster so that
all nodes in the clusters are running only CVM 3.2 or only CVM 3.5,
but never a combination of the two. To do this: