crsctl stat res ora.crf -init -t
== stop
crsctl stop res ora.crf -init
== delete
删除$ORACLE_HOME/crf/db/$HOMENAME/目录中的crfclust.bdb
rm *.bdb
==
start
crsctl start res ora.crf -init
================
问题描述
11gR2的RAC,$GRID_HOME/crf/db/rac1
路径下的crfclust.bd文件非常大,将近20G.
Affects:
Product
(Component)
Oracle
Server (PCW)
Range of
versionsbelievedto
be affected
Versions
BELOW 12.2
Versionsconfirmedas
being affected
Platforms
affected
Generic
(all / most platforms affected)
Fixed:
The fix
for 20186278 is first included in
Description
BDB retension time suddenly change to 1094795585 or some other huge value.
This may cause crfclust.bdb file to grow in size, as crflogd tries to keep
data for longer retention.
Rediscovery
Notes
When hitting this problem, crflogd.log may show the retention size being
suddenly updated to a large size.
------------------------------------------------------------------
2014-02-03 15:36:29.799: [ CRFLOGD][2363541248]updateDBRetention: new retention is 1094795585 seconds and 4702111234474983745 bytes
------------------------------------------------------------------
Current retention value may be checked with following command.
If the retention value (unit: seconds) shows a large value like
1094795585, you are likely to have hit this problem.
$ oclumon manage -get repsize
CHM Repository Size = 1094795585
Workaround
Modify BDBSIZE parameter in config file crf.ora to
adequately lower value. To do so, follow these steps:
1. Issue "$GI_HOME/bin/crsctl stop res ora.crf -init " on all the nodes of cluster.
2. Locate the config file $GRID_HOME/crf/admin/crf.ora
3. Manually edit the crf.ora file on every node of the cluster and
change BDBSIZE tag entry and remove the value (set it to blank) or
set it to a desired value, eg. 61511. Do not delete the BDBSIZE tag itself.
4. Restart ora.crf daemon on every node.
解决方法
该bug在12.2版本修复,变通的解决方法
1.
在集群中的所有节点执行 “GI_HOME/bin/crsctl
stop res ora.crf -init " 2. 跳转到
$GRID_HOME/crs/admin/crf.ora所在目录下
3.
在集群中的所有节点手工编辑crf.ora文件,修改BDBSIZE所在行,移除其所设定的值(设置它为空)
或者 将它的值设定为期望值,例如61511。不要删除BDBSIZE。
4. 在集群中的每个节点重启ora.crf.
参考
Bug 20186278 - crfclust.bdb Becomes Huge Size Due to Sudden
Retention Change (文档 ID 20186278.8)