HP-UNIX系统宕机 .
早上进行db例行巡检的时候发现一个节点2宕机。
首先检查检点2的alter,没有任何发现,而且crs各项资源也offline状态,可以判断应该是系统宕机导致没有任何记录。
通过查看系统登录日志发现有过重启记录:
# last | grep Dec
root pts/1 Mon Dec 17 10:08 still logged in
root pts/0 Mon Dec 17 09:33 still logged in
reboot system boot Sun Dec 16 08:16 still logged in
reboot system boot Sat Dec 15 23:59 - 08:16 (08:16)
但是就是不知道系统重新信息会不会也记录到这里,而且看第3条记录,还still logged in这个只能交给HP工程师来处理了。
检查/etc/shutdownlog发现如下新:
00:03 Sun Dec 16 2012. Reboot after panic: MCA, IIP:0xe0000000008a1a40 IFA:0xc000000006dae000
08:18 Sun Dec 16 2012. Reboot after panic: MCA, IIP:0xe000000000d650a0 IFA:0x20000000777db0cc
检查节点1的alter日志发现如下信息:
Sat Dec 15 23:55:30 2012
Errors in file /opt/oracle/product/admin/xxx/udump/xxx1_ora_4074.trc:
Sat Dec 15 23:55:31 2012
Errors in file /opt/oracle/product/admin/xxx/udump/xxx1_ora_4074.trc:
Sat Dec 15 23:55:34 2012
Reconfiguration started (old inc 100, new inc 102)
List of nodes:
0
检查crs日志如下:
2012-12-15 23:55:16.183
[cssd(4229)]CRS-1612:node xxx2 (0) at 50% heartbeat fatal, eviction in 0.000 seconds
2012-12-15 23:55:23.183
[cssd(4229)]CRS-1611:node xxx2 (0) at 75% heartbeat fatal, eviction in 0.000 seconds
2012-12-15 23:55:24.181
[cssd(4229)]CRS-1611:node xxx2 (0) at 75% heartbeat fatal, eviction in 0.000 seconds
2012-12-15 23:55:28.183
[cssd(4229)]CRS-1610:node xxx2 (0) at 90% heartbeat fatal, eviction in 0.000 seconds
2012-12-15 23:55:29.180
[cssd(4229)]CRS-1610:node xxx2 (0) at 90% heartbeat fatal, eviction in 0.000 seconds
2012-12-15 23:55:30.183
[cssd(4229)]CRS-1610:node xxx2 (0) at 90% heartbeat fatal, eviction in 0.000 seconds
2012-12-15 23:55:30.682
[cssd(4229)]CRS-1607:CSSD evicting node xxx2. Details in /opt/oracle/product/crs/log/xxx1/cssd/ocssd.log.
[cssd(4229)]CRS-1601:CSSD Reconfiguration complete. Active nodes are xxx1 .
检查cssd日志如下:
[ CSSD]2012-12-15 23:55:16.183 [18] >WARNING: clssnmPollingThread: node xxx2 (2) at 50 2.000000e+00artbeat fatal, eviction in 14.489 seconds
[ CSSD]2012-12-15 23:55:16.183 [18] >TRACE: clssnmPollingThread: node xxx2 (2) is impending reconfig, flag 1037, misstime 15511
[ CSSD]2012-12-15 23:55:16.183 [18] >TRACE: clssnmPollingThread: diskTimeout set to (27000)ms impending reconfig status(1)
[ CSSD]2012-12-15 23:55:23.183 [18] >WARNING: clssnmPollingThread: node xxx2 (2) at 75 2.000000e+00artbeat fatal, eviction in 7.489 seconds
[ CSSD]2012-12-15 23:55:24.181 [18] >WARNING: clssnmPollingThread: node xxx2 (2) at 75 2.000000e+00artbeat fatal, eviction in 6.490 seconds
[ CSSD]2012-12-15 23:55:28.183 [18] >WARNING: clssnmPollingThread: node xxx2 (2) at 90 2.000000e+00artbeat fatal, eviction in 2.489 seconds
[ CSSD]2012-12-15 23:55:29.180 [18] >WARNING: clssnmPollingThread: node xxx2 (2) at 90 2.000000e+00artbeat fatal, eviction in 1.491 seconds
[ CSSD]2012-12-15 23:55:30.183 [18] >WARNING: clssnmPollingThread: node xxx2 (2) at 90 2.000000e+00artbeat fatal, eviction in 0.489 seconds
可以获知节点2在这个时刻已经在重新配置集群了,将节点2剔除了集群。
在通过将存储active之后,集群自动在节点2启动并恢复正常生产。
通过/var/adm/syslog/syslog.log 和old日志发现节点系统重启了,奇怪的是竟然没有重启之前的日志信息,只能打包/var/adm/crash目录下的系统crash(可以通过 q4 crash文件大概查看一下)日志信息给HP支持人员。
-The End-