Bacula-users

Re: [Bacula-users] [Bacula-devel] gtraceback : how safe is it to call during normal operations ?

2009-12-01 17:27:53
Subject: Re: [Bacula-users] [Bacula-devel] gtraceback : how safe is it to call during normal operations ?
From: Kern Sibbald <kern AT sibbald DOT com>
To: "Lucas B. Cohen" <lbc AT members.fsf DOT org>
Date: Tue, 1 Dec 2009 23:25:43 +0100
On Tuesday 01 December 2009 19:44:10 Lucas B. Cohen wrote:
> On Tue, 2009-12-01 at 19:17 +0100, Kern Sibbald wrote:
> > > I'm wondering how
> > > safe it is in practice to attach a debugger like gdb or dbx to a
> > > running process to get a stacktrace.
> >
> > As far as I know it is perfectly safe.  I have done so *many* times, and
> > after debugging detached with no harm to running programs.  It is not
> > something I would normally attempt on a "production" program.
> >
> > > Have there been incidents known to have been caused by the use of
> > > btraceback/gdb/dbx during normal operations ?
> >
> > No, not to my knowledge, and as noted below, Bacula never calls
> > btraceback/gdb/dbx during normal operations.
>
> That is how I understand its design from the Problem Resolution Guide
> (PRG). But what I meant as a scenario was more along the lines of a
> repeated, extensive usage by an operator, or as a cron job, or to comply
> with a strict security policy. Or by myself, to try and get a "visual
> feel" for Bacula's architecture with a debugger when I'll begin testing
> and migrating to v3.0.3.
>
> > We do not do live debugging on Bacula.
>
> It's understood that the daemons don't call btraceback upon themselves
> unless it's done by the exception handler. But when btraceback is called
> from the command-line in order to (quote the PRG) "produce an email
> showing you the current state of the daemon [...], and then exit leaving
> Bacula running as if nothing happened. ", does it qualify as live
> debugging ? This is what I was referring to.
>
> The idea is to give a fair warning to uncautious users who would be
> tempted to abuse the feature after having read its manpage. Your comment
> on not normally attempting this in production makes me feel it's
> appropriate.
>
> "Interactive use of btraceback is subject to the usual risks of live
> debugging, which means it can cause Bacula to crash under rare and
> unfortunate circumstances."

I don't know where you came up with the above quote, but though I would not 
recommend using btraceback for "interactive" use, I don't believe the 
conclusion that is drawn is correct.  It goes contrary to everything I wrote 
you in the last email.

>
> Thank you for the details.
>
> > Regards,
> >
> > Kern



------------------------------------------------------------------------------
Join us December 9, 2009 for the Red Hat Virtual Experience,
a free event focused on virtualization and cloud computing. 
Attend in-depth sessions from your desk. Your couch. Anywhere.
http://p.sf.net/sfu/redhat-sfdev2dev
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users

<Prev in Thread] Current Thread [Next in Thread>