Solaris

Hi All,

We a critical application running on the Sun Solaris platform. But, for some reasons the process get's killed abnormally in the production environment and not sure what was causing this. I thought i would use the truss command to trace the errors.. so i used the bellow command

truss -p <pid> once i issued this command my application didn't crash at all. It's been running for 24 hrs with out any issues. previously the application used to crash at least 2 hrs once. Can somebody please explain what's so special when i use the truss on a process id. Did any of you had similar situations ??? Please kindly suggest.


Thank you,
Sudharma.

truss is eating CPU time and maybe this prevents the crash... use "dtrace" for a real-time look at your application. this link is a good startpoint to learn about dtrace:
DTrace Tools

Thanks a lot DukeNuke. I will try this out.


Cheers
Sudharma.

---------- Post updated at 09:22 AM ---------- Previous update was at 09:09 AM ----------

Hi DukeNuke,

When you said it's eating up the CPU time, how does it stop my process being killed. Sorry, just wanted to know how this thing works.

Thank you,
Sudharma.

truss can use a lot of cpu time... i don't know if that is the problem but it CAN be... dtrace in solaris 10+ is the better solution to look for resource (and other) problems.

Thanks DukeNuke


Cheers
Sudharma

Is your application multi-threaded? If so, you're probably dying in a SIGSEGV or SIGBUS and should be able to find a core file somewhere if you've got core files enabled.

If you do find such core files, you've probably got a race condition that running under truss masks. As others have noted, truss takes up some CPU cycles and this can impact the timing of your application in unpredictable ways.

FWIW, this isn't exactly a new type of bug, and I've even seen the exact opposite of your problem, too. An application would run fine until truss was attached to it - a race condition would almost immediately be triggered and the app would crash.