how to figure out where a process hung

Aug 01, 2006 4 Replies

I've got a problem...



I have a process run from cron that hangs in an uninterruptible sleep.



It reads data from a webcam, and writes to a tmpfs partition.



This process runs every 15 minutes; and most of the time it will run just fine, but once in a while, it hangs. Then cron runs it again, and it hangs again. Pretty soon I have a bunch of hung processes that consume all resources, and for all practical purposes my little system is dead.



The frustrating thing is that it happens rarely; the process runs every 15 minutes and sometimes it will run for days just fine, and then it will start hanging.



Is there some way to find out where the process is hung after it is hung up?



The program is spcacat, a very simple snapshot util for webcams using the spca driver: . Anyone have any suggestions? I have 3 weeks to get this up and running, and that doesn't give me much time....


--Yan


o__ ,>/'_ o__ (_)\(_) ,>/'_ o__ Yan Seiner, PE (_)\(_) ,>/'_ o__ Certified Personal Trainer (_)\(_) ,>/'_ o__ Licensed Professional Engineer (_)\(_) ,>/'_ Who says engineers have to be pencil necked geeks? (_)\(_)

What does this mean ? A sleep() call needs to specify a time, so it can't "hang".

Moreover, AFAIK, a user land process only can do uninterruptible sleep (a very short nanosleep() ), if it is assigned very special attributes.

Is it possible that the process waits for some hardware event that does not occur due to defective hardware ?

-Michael

Try to use "strace". When it hangs connect to it with "strace -p " and you will see where it hangs (if it hangs in a system call). If this does not help, try with "ltrace" instead.

Hope it helps Juergen

Hello,

Your cronjob could kill all running instances before starting a new one. That way the ressources would stay free and the system won't get problems. It is not a clean solution, but at least it can keep your ressources free. You could log if any instances are killed (instead of a clean exit) and maybe find some event which causes the program to hang.

That's an idea, but it will only help against the ressource leak of hung processes, not against the problem.

I'd guess the program waits for something (a camera's event?), but never gets it.

Regards, Sebastian

From 'man ps':

PROCESS STATE CODES Here are the different values that the s, stat and state output specifiers (header "STAT" or "S") will display to describe the state of a process. D Uninterruptible sleep (usually IO)

These processes show up as 'D', which means they cannot be killed.

I am guessing that thse processes are waiting for some camera event that never occurs, but I have not figured out why only sometimes....

The camera shares the USB bus with a GPS, which is being polled almost continously. I suspect there is some bus contention which triggers this, but I have no idea where to start looking; all of the code I've looked at looks OK so far.

--Yan

o__ ,>/'_ o__ (_)\(_) ,>/'_ o__ Yan Seiner, PE (_)\(_) ,>/'_ o__ Certified Personal Trainer (_)\(_) ,>/'_ o__ Licensed Professional Engineer (_)\(_) ,>/'_ Who says engineers have to be pencil necked geeks? (_)\(_)

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required