[Cluster-devel] bug reports

[Steven Whitehouse <swhiteho@redhat.com>]

Hi,

On Wed, 2012-10-24 at 13:46 +0200, Heiko Nardmann wrote:
> Am 24.10.2012 13:00, schrieb Steven Whitehouse:
> > Hi,
> >
> > On Wed, 2012-10-24 at 12:38 +0200, Heiko Nardmann wrote:
> >> Hi together!
> >>
> >> Since all (or almost all?) GFS2 developers (as far as I can tell) are
> >> employed by RedHat I wonder whether it makes sense to additionally post
> >> bug reports to this mailing list beside reporting them to the RH support?
> >>
> >>
> >> Kind regards,
> >>
> >>     Heiko
> >>
> > It depends what the reports are really... for those with subscriptions
> > and who are Red Hat customers, then using our official support channels
> > is the best way. That is not to preclude discussing the issue on mailing
> > lists too, if you want to, but by going through our official support
> > channels that does ensure that issues are handled in a timely manner.
> >
> > Using systems like our ticketing system and/or bugzilla means that we
> > have bugs in a filing system where we can keep track of them, and where
> > we can look for common features (sometimes, being able to see several
> > different reports of the same bug, but on different configurations can
> > be very helpful in tracking things down). By contrast issues posted to
> > mailing lists can be more easily lost track of, even though they are
> > likely to reach a wider audience which can be an advantage if the issue
> > is something which another user already has experience of.
> >
> > This particular list however, is intended for development discussion, so
> > thats mostly at the level of proposed patches for both bugs and new
> > features. So if you have a patch which fixes a bug in the upstream code,
> > then do please feel free to post it here whether or not you've opened a
> > bug for it,
> >
> >
> > Steve.
> >
> >
> Hi Steven!
> 
> Since I am busy with the other tasks inside the current project I have
> no time to understand and check the code - so no fixes have to be
> expected from my side. It is just that I am experiencing bugs and maybe
> someone has an idea. Or is experienced enough with the code to almost
> immediately know what might be the reason for it.
> 
> Currently I am in the state that the RH support recommended running the
> debug kernel of RHEL to maybe get further details (I am using 6.1). My
> setup is a two-node cluster (HA) using GFS2 to access a SAN. I have
> tried to run a worst case scenario for GFS2, i.e.
> 
> 1) create traffic on the active node (thus leading to traffic on the SAN)
> 2) run an endless loop of 'find /SAN-Storage -ls' on the passive node
> for some minutes
> 3) stopping the endless loop on the passive node
> 4) unmounting /SAN-Storage on the passive node
> 
> This sequence has lead to a crash of the passive node almost immediately
> after typing 'umount /SAN-Storage' and pressing 'Enter'. First I get the
> following on the console (being logged into the machine using SSH):
> 
> Message from syslogd....
>  kernel:general protection fault: 0000 [#1] SMP
> 
>  kernel:last sysfs file:
> /sys/devices/platform/host8/session2/target8:0:0/8:0:0:1/timeout
> Write failed: Broken pipe
> 
> Then I see a kernel panic on the iDRAC6 console; I've captured a
> screenshot of the stack trace if someone is interested. Sorry, no kdump
> vmcore gets created in this situation.
> 
> Since I am currently considering switching away from GFS2 (being too
> unstable) and instead using ext4 on the SAN (and handle mounting
> explicitly of it on our own) the experienced problems might get lost
> otherwise if not reported here (IMHO).
> 
> 
> Kind regards,
> 
>     Heiko
> 

I assume from the fact that you are in touch with our support team that
you must have opened a ticket. Can you upload that screen shot (if you
have not already) and let me know what the ticket number is?

Steve.