Hello,
> Gesendet: Sonntag, 19. Mai 2013 um 05:34 Uhr
> Von: "David Anderson" <da...@ssl.berkeley.edu>
> An: boinc_dev@ssl.berkeley.edu
> Betreff: Re: [boinc_dev] [patch] mystery solved? Aw: Re: BOINC having too
> many open files - failure in opendir()
>
> As far as I can tell, none of these changes would fix
> a file descriptor leak in the client.
The MFILE class today I read as
class A {
FILE *f;
A::A(string fname) {
f=fopen(fname);
}
A::close{
fclose(f);
}
A::~A() {
// not closing the file pointer
}
}
which leaks a file pointer with every creation of an object of class A that
does not see a close(). For MFILE this is in lib/gui_rpc_client.cpp Please have
a second look at it.... and kindly ignore those 'const' changes ... I did not
notice adding them :o) By grepping through the code I agree that all
instances of the BOINC client that use MFILE with a filename do indeed perform
the close.
I can emperically now confirm that the attached patch does not fix the issue.
When now bringing the boinc client up again, the first thing the client does is
to upload results. The machine is otherwise idle, no idea whom else to blame.
Steffen
> We need a system-call trace that shows open()s.
>
> It's also possible that the system is running out of file descriptors
> because of software other than BOINC.
>
> -- David
>
> On 18-May-2013 4:00 AM, "Steffen Möller" wrote:
> > Dear all,
> >
> > I skimmed through all invocations of (boinc_)?fopen() in api/ and lib/,
> > seeking the respective matching fclose(). What I found missing I placed here
> > http://anonscm.debian.org/gitweb/?p=pkg-boinc/boinc.git;a=blob;f=debian/patches/fopen_closing.patch;hb=HEAD
> >
> >
> as a patch. The trickiest and possibly the most important one is the omission
> of
> a close in the destructor of the MFILE class.
> >
> > Cheers,
> >
> > Steffen
> >
> >
> >> Gesendet: Freitag, 17. Mai 2013 um 22:00 Uhr Von: "Nicolás Alvarez"
> >> <nicolas.alva...@gmail.com> An: "boinc_dev@ssl.berkeley.edu"
> >> <boinc_dev@ssl.berkeley.edu> Betreff: Re: [boinc_dev] BOINC having too many
> >> open files - failure in opendir()
> >>
> >> Get the list of open files (ls -l /proc/$(pidof boinc)/fd) when that
> >> happens. Does the client die after that last fopen() failure? Maybe you
> >> could write a script to log the open file list every few minutes.
> >>
> >> -- Nicolás
> >>
> >> 2013/5/16 "Steffen Möller" <steffen_moel...@gmx.de>:
> >>> Dear all,
> >>>
> >>> every few months I get an error like the one below (taken from the
> >>> stdoutdae.txt) the report too many open files. This is see for about
> >>> three years on several Linux machines, I only recall such with many cores
> >>> (12 or 24), though, Opterons and Xeons alike. Is anything jumping at you
> >>> where to look?
> >>>
> >>> Cheers,
> >>>
> >>> Steffen
> >>>
> >>> 16-May-2013 16:58:33 [World Community Grid] Sending scheduler request: To
> >>> fetch work. 16-May-2013 16:58:33 [World Community Grid] Requesting new
> >>> tasks for CPU 16-May-2013 16:58:36 [World Community Grid] Scheduler
> >>> request completed: got 0 new tasks 16-May-2013 16:58:36 [World Community
> >>> Grid] No tasks sent 16-May-2013 16:58:36 [World Community Grid] No tasks
> >>> are available for The Clean Energy Project - Phase 2 16-May-2013 16:58:36
> >>> [World Community Grid] No tasks are available for the applications you
> >>> have selected. 16-May-2013 16:58:42 [Einstein@Home] Sending scheduler
> >>> request: To fetch work. 16-May-2013 16:58:42 [Einstein@Home] Reporting 4
> >>> completed tasks 16-May-2013 16:58:42 [Einstein@Home] Requesting new tasks
> >>> for CPU 16-May-2013 16:58:46 [Einstein@Home] Scheduler request completed:
> >>> got 1 new tasks 16-May-2013 17:15:53 [Einstein@Home] Sending scheduler
> >>> request: To fetch work. 16-May-2013 17:15:53 [Einstein@Home] Requesting
> >>> new tasks for CPU 16-May-2013 17:15:56 [Einstein@Home] Scheduler request
> >>> completed: got 1 new tasks 16-May-2013 17:30:11 [World Community Grid]
> >>> Can't get task disk usage: opendir() failed 16-May-2013 17:30:11
> >>> [Einstein@Home] Can't get task disk usage: opendir() failed 16-May-2013
> >>> 17:30:11 [Einstein@Home] Can't get task disk usage: opendir() failed
> >>> 16-May-2013 17:30:11 [Einstein@Home] Can't get task disk usage: opendir()
> >>> failed 16-May-2013 17:30:11 [Einstein@Home] Can't get task disk usage:
> >>> opendir() failed 16-May-2013 17:30:11 [Einstein@Home] Can't get task disk
> >>> usage: opendir() failed 16-May-2013 17:30:11 [Einstein@Home] Can't get
> >>> task disk usage: opendir() failed 16-May-2013 17:30:11 [Einstein@Home]
> >>> Can't get task disk usage: opendir() failed 16-May-2013 17:30:11
> >>> [Einstein@Home] Can't get task disk usage: opendir() failed 16-May-2013
> >>> 17:32:31 [Einstein@Home] read_stderr_file(): malloc() failed 16-May-2013
> >>> 17:32:31 [Einstein@Home] Computation for task
> >>> LATeah0024U_80.0_500_-4.66e-10_1 finished 16-May-2013 17:32:31
> >>> [Einstein@Home] md5_file failed for
> >>> projects/einstein.phys.uwm.edu/einstein_S6BucketLVE_1.04_i686-pc-linux-gnu__SSE2:
> >>> fopen() failed 16-May-2013 17:32:31 [---] Can't open
> >>> client_state_next.xml: fopen() failed 16-May-2013 17:32:31 [---] Couldn't
> >>> write state file: fopen() failed; giving up
> >> _______________________________________________ boinc_dev mailing list
> >> boinc_dev@ssl.berkeley.edu
> >> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe,
> >> visit the above URL and (near bottom of page) enter your email address.
> > _______________________________________________ boinc_dev mailing list
> > boinc_dev@ssl.berkeley.edu
> > http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev To unsubscribe,
> > visit the above URL and (near bottom of page) enter your email address.
> >
> _______________________________________________
> boinc_dev mailing list
> boinc_dev@ssl.berkeley.edu
> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
_______________________________________________
boinc_dev mailing list
boinc_dev@ssl.berkeley.edu
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.
