David McDaniel
2007-Jul-19 21:28 UTC
[dtrace-discuss] concurrent tracing of multiple processes
I''ve searched around and cant seem to put my finger on a clear answer.
I have a case where I wish to trace a given function in several processes
simulateously.
So I have a script:
#!/usr/sbin/dtrace -s
pid$1::somefunc:entry { some action; }
pid$2::somefunc:entry { some action; }
pid$3somefunc:entry { some action; }
pid$4:somefunc:entry { some action; }
It works fine with as many as 4 pids, but when I go up to 5, it aborts with an
error message that it is unable to attach to the 5th pid "unexpected system
error".
Is there a specific limit or restriction?
Thanks
--
This message posted from opensolaris.org
Adam Leventhal
2007-Jul-19 23:03 UTC
[dtrace-discuss] concurrent tracing of multiple processes
Working with David offline, we determined that this was caused by file
descriptor exhaustion. He was able to solve the problem by increasing the
limit using ulimit(1) with the -n option.
See this bug:
6277664 Pgrab() failures due to the fd rlimit are not well reported
In the past the kernel group has opined that it could be possible to make
libproc a bit less piggy when it comes to file descriptors. That would also
help ameliorate the problem.
Adam
On Thu, Jul 19, 2007 at 02:28:53PM -0700, David McDaniel
wrote:> I''ve searched around and cant seem to put my finger on a clear
answer. I have a case where I wish to trace a given function in several
processes simulateously.
>
> So I have a script:
> #!/usr/sbin/dtrace -s
> pid$1::somefunc:entry { some action; }
> pid$2::somefunc:entry { some action; }
> pid$3somefunc:entry { some action; }
> pid$4:somefunc:entry { some action; }
>
> It works fine with as many as 4 pids, but when I go up to 5, it aborts
with an error message that it is unable to attach to the 5th pid
"unexpected system error".
> Is there a specific limit or restriction?
>
> Thanks
>
>
> --
> This message posted from opensolaris.org
> _______________________________________________
> dtrace-discuss mailing list
> dtrace-discuss at opensolaris.org
--
Adam Leventhal, Solaris Kernel Development http://blogs.sun.com/ahl