SUNSOLVE EARLYNOTIFIER(SM) ALERT

SunSolve EarlyNotifier Alert is published periodically to provide 
Sun contract customers with the latest and most important technical 
information regarding Sun hardware and software.

***************************************************************************

Applies to users of all RSM2000, A3000, A3500 (Sonoma) and SC2.1 
(SunCluster 2.1) sites.

PROBLEM DESCRIPTION

RDAC retry by default is set to 8, which means when there is some sort
of I/O failure thus forcing RDAC to try the alternate path. For every I/O
that gets queued up, it will try that I/O 8 times before giving up.

Example    Mirroring an A3500 to another A3500 with Volume Manager.
Scenario:  Using Oracle as the backend database.
		  
	   If the primary A3500 were to lose all power, the plex would
	   not detach due to RDAC not sending appropriate messages back
	   to sd. This means that since the plex did not detach and move
	   over and continue on the other A3500, the front-end application
	   to Oracle would time out.

WORKAROUND

The attached script to modify RDAC retry count needs to be run on each node
in a 2-node cluster within Sun Cluster 2.1. No rebooting is needed after
running the script.

The script modifies the rdriver object file so that the RDAC retry count is
set to 1.  Meaning that for every I/O that is queued up, it will retry it
once before RDAC fails, and in turn pass the appropriate messages onto sd.

SOLUTION

Patch 106707-01 is being tested and approved for distribution through
SunSolve Online.  Once the patch is released, there will be an environment
variable that will be set in the rmparams file and the script will then be
removed from the SunSolve website.


-------begin script-----------

#!/bin/sh
adb -kw << "EOF" >/dev/null 2>&1
Cf_MaxStratTries?W 1
$q
EOF

--------end script------------

from Ray Tully (tully@bast)