Tuning the Cluster Interconnect - Expert Oracle RAC Performance Diagnostics and Tuning - page 474

Script: MVRACPDnTap_raclostblocks.sql
SELECT
A1.INST_ID ,
A3.INSTANCE_NAME,
A3.STARTUP_TIME,
ROUND (SYSDATE - STARTUP_TIME) "Days",
A1.VALUE ,
A2.VALUE
FROM GV$SYSSTAT A1,
GV$SYSSTAT A2,
GV$INSTANCE A3
WHERE A1.NAME = 'gc blocks lost'
AND A2.NAME = 'gc blocks corrupt'
AND A1.INST_ID = A2.INST_ID
AND A1.INST_ID = A3.INSTANCE_NUMBER
ORDER BY 1;
Instance Instance Blocks Blocks
Instance Name Startup Time Up Days Lost Corrupt
-------- ---------- ------------------ -------- -------- --------
1 MVP1 11-OCT-10 12:55:38 1 34263 0
2 MVP2 11-OCT-10 12:56:27 1 30413 0
3 MVP3 11-OCT-10 12:56:33 1 2265 0
4 MVP4 11-OCT-10 12:56:43 1 137 0
5 MVP5 11-OCT-10 12:51:23 1 18886 0
To solve the lost blocks issue, the following troubleshooting steps were taken:
Step 1
The first step is to understand if the cluster interconnect was configured correctly. Did the database really use the
correct interconnect? A dump of the OCR file was taken using the following command:
$GRID_HOME/bin/ocrdump < output file name>
$GRID_HOME/bin/ocrdump /home/oracle/mvallath/ocrdumpIC11oct10.dmp
Step 2
The private network alias (highlighted following) was rightly being referenced. For example, the OCR file output
(following) showed the correct entries:
[SYSTEM.css.node_numbers.node1.nodename]
ORATEXT : oradb3
SECURITY : {USER_PERMISSION : PROCR_ALL_ACCESS, GROUP_PERMISSION : PROCR_READ, OTHER_PERMISSION :
PROCR_READ, USER_NAME : root, GROUP_NAME : root}
[SYSTEM.css.node_numbers.node1.privatename]
ORATEXT : oradb3-priv
SECURITY : {USER_PERMISSION : PROCR_ALL_ACCESS, GROUP_PERMISSION : PROCR_READ, OTHER_PERMISSION :
PROCR_READ, USER_NAME : root, GROUP_NAME : root}