[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [leafnode-list] leafnode hanging

On 19 Feb,  
     Matthias Andree <matthias.andree@xxxxxx> wrote:

> Brian D schrieb am 2004-02-19:
> > I installed Leafnode 1.9.50 today, and so far it has hung twice when
> > reading from a dodgy server (text.news.ntlworld.com) which I only use for
> > a few newsgroups. This prevents it from reading from the other servers at
> > the scheduled next run (10 minutes). It has happenned on earlier
> > versions, so it may be coincidence. Logging was set to 0 so not much to
> > show there. 
> > 
> > Is there a timeout to set to abort an inactive fetch? or should I
> > automate my script to kill the process before calling fetchnews. Any
> > suggestions?  
> Does it hang with 100% CPU or with 0% CPU?


> I'd like to analyze this first, if that's ok with you:
> If you've installed with "make install-strip", please reinstall with "make
> install". If you've installed from RPM, skip the backtrace steps for now,
> they won't produce any useful information.

I installed using rpm -tb followed by rpm -U. I have now installed 1.9.51 as
I think it could be the slow download of the groupinfo file.

> If you have used "make install" to install, please
> 1. debugmode=3 in /etc/leafnode/config and modify your /etc/syslog.conf
>    and restart syslogd according to instructions in the README file so
>    we have more logging.  If it then hangs, please obtain a stack
>    backtrace (see the FAQ for instructions).
> 2. use your shell's command to increase the core file size limit,
>    bash uses for instance "ulimit -Sc unlimited"
> 3. run fetchnews -vvv and wait until it hangs
> 4. press Ctrl-\   - this should print "Quit (core dumped)"
> 5. now obtain the core file as described in the FAQ.
> If the backtrace looks "all numbers" and GDB complains "no symbols found"
> (which is the case after make install-strip, with "-s" in LDFLAGS or CFLAGS
> and for RPM installs), the backtrace is worthless unfortunately.
> Anyways, please bzip2 or gzip the syslog output and send it to my private
> mail address (not to the list) or stuff them on your FTP server and tell me
> the URL. If the backtrace has symbolic names, paste it into the mail or
> attach it uncompressed (plain text).
> If you don't have a backtrace, try running fetchnews under supervision of a
> syscall tracer, such as strace, truss, trace, ktrace, ltrace -S and gzip
> the output and make it available to me or attach it to a mail you send to
> my private email address.

I'll save this info until I check 1.9.51, and I've sobered up.

  Brian Duffell    RiscPC600      | StrongARM 233Mhz RISC OS 4 66Mb
  Linux PCs on RISC OS network    | <brian@xxxxxxxxxxxxxxxxxxxxxxxx> 
  Darlington Dolphin Masters ASC  | <www.darlington-masters.org>

leafnode-list mailing list