logging of cluster

[khilari]

Hi guys i wanted to know how many different ways i can check the logging of the cluster. Linke i can check in syslog.log. Then in the control file log, where else. And can i start a particular logging service???
Thanks

[Mridul Shrivastava]

Hi,

There are ways to increase the logging in cluster using cmsetlog and making changes to inetd.conf file.

for cmsetlog and other options detail refer the following thread:

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1090221&admit=109447626+1200691107249+28353475

[Mridul Shrivastava]

you can change serviceguard log level (in syslog.log) with the command:
#cmsetlog 2
where 2 is the level (0 by default), you canincrease the level with 3, 4 ... I think 2 is a good value after you'll have to many messages in syslog.log. (answer yes 2 times to the warning questions).
After debug step, use command:
#cmsetlog -r
to return to default level.

by using -f option you can redirect the output to another file so that file will have all the messages related to MCSG only.

I remember there was an option -T which can be inserted in the hacl-cfg line for udp/tcp, however that is used for enahanced logging which is normally not required as that may create MBs of data in hours/day.

[John Bigg]

In terms of normal logging it appears you already know about the standard logs to check, i.e. syslog and package logs. There are no other log files directly associated with the cluster although some of the other daemons such as the cmsnmpd and cmomd have separate log files.

With regards to the responses about turning up logging this is intended only for troubleshooting and is unsupported to use during normal production.

Firstly, the method for doing this varies from release to release and has changed dramatically in 11.18 for the commands.

Secondly, turning up logging risks the integrity of the cluster. Firstly if you have logging turned up writing to a log file on disk and you have a disk failure then the process can hang and you can loose the entire node. Secondly, logging can affect timing and can cause communications problems if you are not aware of this and do not know what you are doing.

Therefore I would not suggest you turn up logging using cmsetlog or any other utility or option unless you are instructed to do so by HP support. I doubt this information would help you anyway and often leads to more concern about possible problems than it helps. At higher log levels all sorts of errors become visible which are not real errors but are expected. For example cmclconfd will report socket errors if the cluster is not running as it attempts to communicate with cmcld which would not be running.

I'd stick with the log files you already found and not worry about increasing default log levels.