Specifications

3-4 Sun StorEdge A1000 and A3x00/A3500FC Best Practices Guide • November 2002
3.1.5 Multi-Initiator/Clustering Environment
■ Sun Cluster is the only clustering/multi-initiator environment tested and verified
by Sun with the A3x00. A number of parameters should be modified to run an
A3x00 under the Sun Cluster 2.1/Sun Cluster 2.2 environment. See
Rdac_RetryCount, Rdac_NoAltOffline and Rdac_Fail_Flag in the RAID
Manager 6.1.1_u1 or 6.1.1_u2 patch number 106707-03 or later.
■ See the Sun Cluster documentation for specific Sun Cluster requirements.
For Sun Cluster 2.1, refer to the following web site:
http://suncluster.eng.sun.com/engineering
For Sun Cluster 2.2, refer to the following web site:
http://suncluster.eng.sun.com/engineering/SC2.2/fcs_docs/fcs_d
ocs.html
For Sun Cluster 3.x, refer to the following web site:
http://suncluster.eng.sun.com/products/SC3.0/
■ Running rm6 diagnostic commands in a multi-host configuration
When rm6 diagnostic commands (e.g. healthck, drivutil) are run, lock files
are used on the host to serialize execution and protect against these commands
being run simultaneously against an A1000/A3x00/A3500FC module.
This design means that running multiple rm6 diagnostic commands at the same
time, on the same host, is safe.
Running rm6 diagnostic commands from more than one host connected to a
shared A1000/A3x00/A3500FC at the same time, however, can cause controllers
to go offline, potentially affecting data availability.
This situation can occur when running explorers or third party application
packages (e.g. BMC Patrol) from multiple hosts with shared
A1000/A3x00/A3500FC storage at the same time.
Additionally, Sun Remote Services software runs rm6 utilities via the scripts
healthck.sh and drivutil.sh (included in the package SRSsymod).
In all of these cases, it’s important to modify procedures or configurations to
ensure rm6 diagnostic commands are never executed from more than one host
attached to a shared A1000/A3x00/A3500FC at the same time.
Running the rm6 GUI from more than one host attached to a shared
A1000/A3x00/A3500FC at the same time should also be avoided. Even when the
administrator has not actively executed diagnostic commands via the GUI, an idle
rm6 GUI may nevertheless be probing the A1000/A3x00/A3500FC. The customer
will still be exposed to the same problem of controllers going offline.