[collectd] Segfaults with collectd 4.0.5 on RHEL 4 U5 x86_64

Ulf Zimmermann ulf at atc-onlane.com
Tue Aug 7 00:05:45 CEST 2007

> -----Original Message-----
> From: Florian Forster [mailto:octo at verplant.org]
> Sent: 08/06/2007 14:30
> To: Ulf Zimmermann
> Cc: collectd at verplant.org
> Subject: Re: [collectd] Segfaults with collectd 4.0.5 on RHEL 4 U5
> Hi Ulf,
> On Mon, Aug 06, 2007 at 11:28:37AM -0700, Ulf Zimmermann wrote:
> > I had two systems where collected died this weekend.
> do I understand that right that it did work for some time? Those bugs
> are of the especially nasty kind :/

It was rolled out Sunday July 29, machines got rebooted again Monday, so
about 4-5 days of uptime.

> Since this typo of problem is very hard to reproduce, the best change
> you've got is to (re)build collectd with debugging symbols enabled
> (./configure --enable-debug $OTHER_OPTS), enable the creation of a
> corefile (ulimit -c unlimited, or similar, depending on your shell)
> ``hope'' that the issue hits once more.

I need to build a new rpm later, right now still busy with a whole lot
of things from our Oracle server upgrades to EL4 64bit.

> > Aug  4 04:11:09 dbprd01 kernel: collectd[8452]: segfault at
> > 0000000000000000 rip 0000002a993346d0 rsp 0000000042802300 error 4
> This isn't of much use, I'm afraid.. What we need to know is the
> position in the code, not the position in memory. The debugging
> will provide the neccesary glue here. If you have a binary with
> debugging symbols and you have a corresponding corefile, the bug
> get it's warm underwear ;)
> > My version of collectd only has a limited number of plugins compiled
> > and in the rpm, which are cpu, csv (actually going to take that one
> > out too), df, disk, exec, interfaces, iptables, irq, load, logfile,
> > memory, network, perl, processes, rrdtool, swap, syslog and
> Do you use all of these plugins? If so, could you give me a short
> feedback what you use them for, especially the `exec', `perl' and
> `unixsock' plugins?
> Plugins you don't use right now can be disabled without losing
> functionality. But since they're not actually used it's unlikely that
> that you can get rid of the bug in that way..
> If you need any help let me know, or visit us in IRC :)

I don't actually need exec, perl and unixsock, I think, just my rpm
build turns off certain modules and as these got added recently I
haven't specific turned them off. I original built my own rpm because of
not wanting the requirement of MySQL libs to be installed. So I went and
turned off any modules I don't care about.

> Regards,
> -octo
> --
> Florian octo Forster
> Hacker in training
> GnuPG: 0x91523C3D
> http://verplant.org/

Regards, Ulf.

ATC-Onlane Inc., T: 650-532-6382, F: 650-532-6441
4600 Bohannon Drive, Suite 100, Menlo Park, CA 94025

More information about the collectd mailing list