怎樣確認我的cluster 是完全正常的呢?

[watermelonyu]

我的環境是oracle rac 9207

on tru64 5.1b cluster最新的patch

最近發生一件事,兩台oracle 都hung住

連sqlplus 都進不去

本來懷疑是cluster 有狀況,但從log 或者

caa_stat,clu_quorum status 看都是正常的

不知道是否有更detail 的東西可以看cluster的狀態(因為oracle 完全沒有產生任何log可以看),所以想先從cluster 先下手

[watermelonyu]

1. 在兩台執行 top, 看看前幾名 process 是否為 oracle lmd 等 RAC 程式?



2. 透過 tcpdump 瞭解 hang 住時, memory channel 是否有任何 packet 在溝通

[watermelonyu]

可否解釋一下

tcpdump 如何操作

因為現在機器不在我身邊

[watermelonyu]

Hi Joey,



Does it has encountered system processes hung while DB hanging?



1) Please enable the collect process running and call HP for further analysis if hang again.

# mkdir /usr/field/collect_log

# collect -n -S50 -i 120 -H d0:5,1m -W 1h -f /usr/field/collect_log/`hostname`



2) Tuning some system parameters in /etc/sysconfigtab for RAC requirement. Such as:



rdg:

max_objs = 8096

msg_size = 32768

max_async_req = 2048

max_sessions = 2048

rdg_max_auto_msg_wires = 0



See the related informations from:

http://www.oracle-base.com/articles/9i/Oracle9iRACInstallationOnTru64.php



http://www.dbis.informatik.uni-goettingen.de/Teaching/oracle-doc/admin-guide/appd_tru.htm



http://paaf.gov.kw/elib/paaf-full-installation.pdf



Best regards,

Richard.

[watermelonyu]

Hello,Eric

你遇到這種hung 住的問題

都怎樣處理的

[watermelonyu]

Hi Joey,



For example for NFS collection via tcpdump on GEb "alt0" interface as below:

# cd /dev/; ./MAKEDEV pfilt

# pfconfig +copyall alt0

# tcpdump -i alt0 -w /usr/tmp/nfs_write.dmp -s300 udp port 2049



Collect all of packets on tu0 get from/to remote host alpha_node via tcpdump:

# tcpdump -i tu0 dst host alpha_node



Best regards,

Richard.

[watermelonyu]

Hi Joey,



> 你遇到這種hung 住的問題都怎樣處理的

Please running "dumpsys" if system is still running.



Otherwise, forcing a HW crash dump via button then

execute the following command on console.

>>> crash

>>> boot

..

..

# sys_check -escalate



Please collect the /var/tmp/escalate.tar and the latest crash dump files on /var/adm/crash/ directory if necessary!



PS: Call HP for further analaysis or

call 3rd party if it is an AP issue.



Best regards,

Richard.