Cleaning up huge crfclust.bdb in Grid home

Details:

Problem:	The Home directory of User grid become huge
Path:	/app/oracle/11.2.0.4/grid/crf/db/db-node01
Key Word:	CRF --> CHM(Cluster Health Monitor)
Files:	one *.bdb(crfclust.bdb) file's size seems to be more than 88GB.

Reason:

I found that it seems like bug 12711827.
You can apply Patch 10165314 to fix this issue Or NOT.
As quick fix, you can Clear them manual.

Resolution:

Plan A: Clear it manual.

grid@db-node01:~$ crsctl stat res -init -t

1. --------------------------------------------------------------------------------
2. NAME            TARGET    STATE      SERVER                   STATE_DETAILS
3. --------------------------------------------------------------------------------
4. Cluster Resources
5. --------------------------------------------------------------------------------
6. ora.asm
7.     1           ONLINE    ONLINE     db-node01                Started
8. ora.cluster_interconnect.haip
9.     1           ONLINE    ONLINE     db-node01
10. ora.crf
11.     1           ONLINE    ONLINE     db-node01
12. ora.crsd
13.     1           ONLINE    ONLINE     db-node01
14. ora.cssd
15.     1           ONLINE    ONLINE     db-node01
16. ora.cssdmonitor
17.     1           ONLINE    ONLINE     db-node01
18. ora.ctssd
19.     1           ONLINE    ONLINE     db-node01                OBSERVER
20. ora.diskmon
21.     1           OFFLINE   OFFLINE
22. ora.drivers.acfs
23.     1           ONLINE    ONLINE     db-node01
24. ora.evmd
25.     1           ONLINE    ONLINE     db-node01
26. ora.gipcd
27.     1           ONLINE    ONLINE     db-node01
28. ora.gpnpd
29.     1           ONLINE    ONLINE     db-node01
30. ora.mdnsd
31.     1           ONLINE    ONLINE     db-node01
32. ---------------------------------------------------------------------------------

grid@db-node01:~$ crsctl stop res ora.crf -init

grid@db-node01:~$ crsctl stat res ora.crf -init -t

grid@db-node01:~$ oclumon manage -get reppath

1. 
   
2. CHM Repository Path = /app/oracle/11.2.0.4/grid/crf/db/db-node01
3.  
4.  Done

root@db-node01:~# cd /app/oracle/11.2.0.4/grid/crf/db/db-node01

root@db-node01:/app/oracle/11.2.0.4/grid/crf/db/db-node01# rm -f *.bdb

grid@db-node01:~$ crsctl start res ora.crf -init  grid@db-node01:~$ crsctl stat res ora.crf -init -t

Plan B: Maybe, we can clear it automatically.

db-node01:/app/oracle/11.2.0.4/grid/crf/db/db-node01$ oclumon manage -get repsize

1. 
   
2. CHM Repository Size = 1931492640  // It is the default value, Maybe this is the bug maker. It's a time(second).
3.                                   // This 'Size' is not about the *.bdb files' size, It's how often (s) to resize.
4.  Done

grid@db-node01:/app/oracle/11.2.0.4/grid/crf/db/db-node01$ oclumon manage -repos resize 172800

//  2 days * 24 hours * 60 minutes * 60 seconds

1. db-node01 --> retention check successful
2. db-node02 --> retention check successful
3. New retention is 172800 and will use 3719208960 bytes of disk space
5. CRS-9115-Cluster Health Monitor repository size change completed on all nodes.
7. Done

grid@db-node01:/app/oracle/11.2.0.4/grid/crf/db/db-node01$ oclumon manage -get repsize

1. 
   
2. CHM Repository Size = 172800
   
3. 
   
4.  Done

Now what?
Wait, for 2 days!
If Plan B really works, you won't have to apply the patch!

Moitoring Plan:
repsize_mon.sh:
#!/usr/bin/sh

echo `date -u "%Y%m%d"`' | '`/usr/bin/du -sk /app/oracle/11.2.0.4/grid/crf/db/db-xxx01`

crontab -l:
# Monitoring the bdb directory.
0 22 * * * /usr/bin/sh /home/oracle/cjs/repsize_mon.sh >> /home/oracle/cjs/bdb_dsk_usage.txt 2>&1

    After check the bdb_dsk_usage.txt at one month later, I found the resize is always between 600M to 630M.
It really means that Plan B did works!!!

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/25521690/viewspace-2122284/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/25521690/viewspace-2122284/