PROBLEM:	(QAR 49188)	(Patch ID: TCR141-006)   
********
This fixes a problem with the Cluster Monitor (cmon) in which ASE service 
status and node UP/DOWN status display may have been incomplete or incorrect.  
It is caused by asymmetric connections between tract daemons where all daemons 
are not bi-directionally connected to each other.  This causes ASE status 
reports to flow asymmetrically between cluster nodes such that the displays on 
two different instances of the Cluster Monitor on two nodes in the cluster show 
different information.

Careful examination of the user.log file in the syslog hierarchy on each 
cluster node will often reveal that all tract daemons are not interconnected 
with all other tract daemons.  The connect messages are printed at system boot 
time and so appear near the top of the user.log file.

This condition can always be cleared by killing and restarting the tract daemons
and Cluster Monitor (cmon) programs, then restarting the submon daemons, then 
after 60 seconds restarting the cmon programs.

A fairly consistent way to reproduce this problem is to halt a cluster node 
from multi-user operation using the halt button, then reboot it.  This will 
normally cause one or more cluster nodes to have asymmetric connections to the 
rebooted node's tract daemon.  Again, this condition is cleared by the restart 
procedure described in the installation instructions.

PROBLEM:	(QAR 44746, QAR 39919)	(Patch ID: TCR141-006)
*******
These QARs both report the same problem: a complete depletion of system socket 
resources, the result of tractd daemons doing repeated connect retries.  This 
problem is seen most commonly when all nodes in a three- or four-node cluster 
are booted simultaneously and is also fixed by the patch described as fixing 
problem QAR 49188.

The socket depletion problem is noticed when socket-based operations such as 
rsh and rlogin fail on the afflicted host.  In these circumstances the user.log 
file contains many (10's or more) tract daemon interconnection retry messages.

This condition can always be cleared by killing and restarting the tract daemons
and Cluster Monitor (cmon) programs, then restarting the submon daemons, then 
after 60 seconds restarting the cmon programs.