dear all:

"system crash" means that:
(1)I can ping it , but can not connect to it (via ssh or http or anything else)
(2)the major role of that machine is to run a web application written in java, every 3 weeks(almost that), the users inform us and say they can log in that web application , and we find the situation just like (1) says.

When this happen , we can only restart the machine by pushing the power button.(It's sad !)

I've try to read all the log files located in /var/log but find nothing interesting.

Can anyone give me some advices : what should I do next? what investigations I should try ?

Thanks in advance.

Ken @ Taiwan ~

Moved: This thread is more suitable in the Linux Server forum and has been moved accordingly to help your thread/question get the exposure it deserves.

- What hardware, amount of RAM, OS, release, kernel?
- What Java application and what memory settings?
- When did this start and what happened at that time hardware, software or configuration-wise?
- Wat (debug) logging and system statistics data (SAR) is available?

>>What hardware, amount of RAM, OS, release, kernel?

It's a HP Proliant ML-380 with 4GB ram running CentOS 5.4

>>What Java application and what memory settings?

-Xms512m -Xmx2048m -XX:MaxPermSize=256m

>>When did this start and what happened at that time hardware, software or configuration-wise?

The system is up and running since April, 2010. The first time it crash like this is at November, 2010.
No any special event happens, It just suddenly turn to be unresponsive.

>>Wat (debug) logging and system statistics data (SAR) is available?

Log files produced by syslogd + (cpu loading/JVM Heap space info.) collected by our homemade scripts.

- What Java application, version, release, vendor?
- If homebrewn is there a developer who knows his way around log4j to provide you with more war / garbage collection / memory / debug logging?
- If bought did you contact the vendor?


- No software updates or re-configuration at all?


And what errors do those logs show before the machine becomes unresponsive?
And what do the load / heap logs show wrt memory usage?

While I'm no proponent of doing things this way would regularly restarting Jakarta (when there's no users connected and say once per month?) help keep the machine on its feet longer while you fix the issue?

paste details from the log file so that we can help.

Thanks for everyone trying to help me ...
But my problem is : I am not sure what logs should I read and provide its content here .

The cpu loading and jvm heap consumption logs made by our own have no change before the system crash.

And the system logs in /var/log produced by syslogd seems ok to me after I read some of them like secure, message, etc.