Veritas Storage Foundation 5.1 SP1 for Oracle RAC Administrator"s Guide (5900-1512, April 2011)

Troubleshooting VCSIPC
This section discusses troubleshooting VCSIPC problems.
VCSIPC wait warning messages in Oracle trace/log files
When Gigabit Ethernet interconnections are used, a high load can cause LMX/LLT
to flow-control VCSIPC, resulting in warning messages to be reported in the Oracle
trace file. The default location for the trace file is $ORACLE_HOME/rdbms/log;
it may have changed if the parameters background_dump_dest or user_dump_dest
have been changed. The messages resemble:
.
Unix process pid; 9560, image: oracle@MCB4800 (LMS0)
*** 2003-03-22 10:18:46.370
*** SESSION ID:(5.1) 2003-03-22 10:18:44.387
VCSIPC wait: WARNING: excessive poll done, 1001 times
VCSIPC wait: WARNING: excessive poll done, 1001 times
VCSIPC wait: WARNING: excessive poll done, 1002 times
VCSIPC wait: WARNING: excessive poll done, 1003 times
VCSIPC wait: WARNING: excessive poll done, 1004 times
VCSIPC wait: WARNING: excessive poll done, 1005 times
.
As a workaround, you can change the LLT lowwater mark, highwater mark, and
window values for flow control. Please contact Veritas support for more
information about changing these values.
VCSIPC errors in Oracle trace/log files
If you see any VCSIPC errors in the Oracle trace/log files, check the
/var/adm/syslog/syslog.log file for any LMX error messages.
If you see messages that contain any of the following:
. . . out of buffers
. . . out of ports
. . . no minors available
See “About LMX tunable parameters” on page 258.
If you see any VCSIPC warning messages in Oracle trace/log files that resemble:
connection invalid
Troubleshooting SF Oracle RAC
Troubleshooting VCSIPC
224