Crash report....still not working

Ben Escoto
Sun, 28 Apr 2002 22:44:11 -0700

Content-Type: text/plain; charset=us-ascii

>>>>> "KS" == Kevin Spicer <Spicer>
>>>>> wrote the following on Fri, 26 Apr 2002 09:44:17 +0100

  KS> I believe that I have now fixed this - it appears to have been
  KS> caused by a limit of 64 on the maximum number of file
  KS> descriptors.  Fixed it by adding ulimit -n unlimited to the
  KS> start of my cron job.
Well, I'm glad it works now for you, but I still don't see why
rdiff-backup would have opened more than 64 files simultaneously.

  KS> etc...  I still have no idea what is causing this (although it
  KS> might be helpful to know that the directories I'm backing up
  KS> have lots and lots of small files (e.g. many at 50 / 60 bytes)

rdiff-backup seems to handle these situations ok.  Maybe it is a
Solaris thing, though.

  KS> open(/dev/zero): Too many open files libthread panic:
  KS> alloc_chunk (PID: 4565 LWP 1)

Why does the error always happen on /dev/zero?  This can't be
coincidence?  Device file handling hasn't been well tested on
non-linux systems, so if the problem is with /dev/zero maybe
--exclude-device-files would help (or --exclude /dev/zero :-)).

    It seems you've also encountered a bug when trying to resume a
previous backup?  I will look into it..

Ben Escoto

Content-Type: application/pgp-signature

Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Exmh version 2.5 01/15/2001