Patch-ID# 103259-05
Keywords: mond clustmon dlm sparc cluster 
Synopsis: Sparc Cluster PDB 1.1 jumbo patch
Date: Mar/28/97

Solaris Release: 2.4

SunOS Release: 5.4

Unbundled Product: Sparc Cluster PDB

Unbundled Release: 1.1

Relevant Architectures: sparc

BugId's fixed with this patch: 1236816 1234156 1256619 4039447

Changes incorporated in this version: 4039447

Patches accumulated and obsoleted by this patch: 

Patches which conflict with this patch: 

Patches required with this patch: 

Obsoleted by: 

Files included with this patch: 

	kernel/drv/dlm
	kernel/drv/dlm.conf
	usr/sbin/in.mond
	opt/SUNWcluster/bin/clustmon
	opt/SUNWcluster/bin/reconf_ener
	etc/opt/SUNWcluster/conf/TEMPLATE.cdb


Problem Description: 

	4039447 (SUNWdlm)

	Testing 1.1 latest patch with dlmstress tests 
	resulted in panic and sometimes hang in 
	dlm_enable_processes().

	Putting a QUEUE_DELETE_RESET before the last 
	qwait() in implicit_detach() and also flushing 
	the write_queue before the last for(;;) loop.


	1256619 (SUNWpdb)

	Both nodes will abort if the communication between them is lost
	while a reconfiguration is taking place. (fatal step 2 time out)
	This also requires a change to the cdb file, which the postinstall
	script supplied with this patch will take care of it for you.

 	(SUNWclmon)
 
	clustmon updated to work with the updated revision of mond 
	(v1.1 r1.8) in this patch.

	(SUNWdlm)

	With large number of Oracle processes the effective throughput of 
	DLM lock transactions would become very low essentially making the 
	DLM a performance bottleneck. This change removes this performance 
	bottleneck by reworking some of the code which would delay DLM 
	processing in lieu of qwriter() requests by deadlock detection code.

	Under high load and high memory pressure there was a 1 instruction
	window where a resource being used by the DLM could be accessed after
	being freed.  This would cause a crash.

	Also, fixed a problem where if a Double BASTs from the dlm could cause 
	a potential hang of the Oracle database: 

	WARNING: ID[SUNWcluster.dlm.3100] double BAST from master, lock=11421b7c

	1234156 (SUNWmond)
	
	mond's logreader does not handle a change in the year correctly. 
	Here is an example:

	0. a syslog message file on a node from '95
	1. year becomes '96
	2. start up clustmon against this node
	3. all the messages in the syslog file from '95 shows up in message viewer as
   	'96.

	1236816 (SUNWdlm)

	Oracle will hang on both pdb nodes if one node runs out of swap space. 


Patch Installation Instructions: 
-------------------------------- 
Refer to the Install.info file for instructions on using the
generic 'installpatch' and 'backoutpatch' scripts provided with
each patch.  Any other special or non-generic installation
instructions should be described below as special instructions.

Special Install Instructions: 
----------------------------- 

   This patch should be installed with both Oracle and PDB stopped.
   The system should be rebooted after you install the patch.