HP Tru64 UNIX and TruCluster Server Version 5.1.B-4 Patch Summary and Release Notes (13156)

• Fixes a tok_wait hang problem on Sierra Clusters.
• Adds the ability to change the default interconnect interface name.
• Corrects several problems in the cluster install and upgrade utilities.
• Fixes a problem in which an RDG (Reliable DataGram) kernel thread can starve other
timeshare threads on a uniprocessor cluster member. In particular, system services such as
networking threads can be affected.
• Fixes minor issues with cfsstat command-line options and return values.
• Prevents panics seen with cluster server-only (for example, MFS) mounts.
• Fixes a condition that causes the panic pg_nwriters going negative when ubc_page_release()
is called from cfs_getpage().
• Corrects a problem in the RDG component in which multiple Oracle instances are unable
to be properly configured when using RDG over a LAN rather than Memory Channel.
• Provides a sticky connection feature for a cluster alias.
• Updates sysconfig to use the cluster interconnect, allowing for a greater SSI collaboration.
This will help with changing variables on hung systems, single user systems, and normal
running systems.
• Improves device error processing in drd.
• Corrects a boot hang problem seen on large-scale Sierra Cluster configurations caused by a
missed wake up in the kernel group services code.
• Alters the behavior of the cluster NFS client with TCP mounts so that when a remote server
is down, the cluster NFS client will use nonreserved ports to see if the remote server is up.
• Introduces a new CFS tunable attribute that may benefit the performance of client reads of
clone files under certain circumstances.
• Addresses an assertion caused by a bad user pointer passed to the kernel via sys_call.
• Corrects a condition that results in excessive context switching and CPU load due to a heavy
use of the cluster alias on large SMP and NUMA machines .
• Enhances /sbin/advfs/tag2name to print out the name of the associated directory, given the
tag of an index file.
• Increases performance scalability and extends the reliability of the Internode Communications
Subsystem in a cluster configured with Memory Channel as the cluster interconnect.
• Improves detection of possible race conditions during CFS recovery.
• Adds a cluster panic facility to the kernel.
• Addresses the following:
— An issue in which new ICS server daemons and handles are created one at a time each
time the low water mark for each is reached, thereby causing a nanny daemon to be
called more frequently than it needs to.
— An issue in which no mechanism exists for the user to adjust the high and low water
marks for ICS free handles, which can result in poor performance during rapidly
increasing loads.
• Fixes a problem in which cluster alias connections are not distributed among cluster members
according to the defined selection weight.
• Fixes a memory leak in the cluster alias subsystem.
• Fixes an issue with ICS (Internode Communication Services) on a NUMA-based system in
a cluster.
• Fixes a problem in the cluster kernel in which a cluster member panics while doing remote
I/O over the interconnect.
• Fixes a hang that occurs when multiple nodes are shutting down simultaneously; fixes a
Cluster File System panic that occurs when using raw Asynchronous I/O; and provides
additional code to assist in problem diagnosis.
4.2 Summary of TruCluster Server Software Patches 161