[ofa-general] Re: osm error messages

Michael S. Tsirkin mst at dev.mellanox.co.il
Sat Mar 24 10:24:21 PDT 2007


> Quoting Hal Rosenstock <halr at voltaire.com>:
> Subject: Re: osm error messages
> 
> On Fri, 2007-03-23 at 12:41, Douglas Fuller wrote:
> > On /21/07 2:53 PM, "Hal Rosenstock" halr at voltaire.cm> wrote:
> > 
> > > On Wed, 200-03-21 at 13:29, Douglas Fulle wrote:
> > >> I'm seeing some sporadic error activity from OpnSM (FED 1.1; osm.log>> below) that ay correlate with some ob failures -- I'mtrying to get to the
> > >> bottom of this.
> > >> 
> > >> efore seeing this, I isolatedand disabled with ibortstat what ppeared
> > >> to be a ba intenal port n one of our core switches.  That leads me to
> > >> suspectI have a switchmisbehaving somwhere.
> > >> 
> > >> ithout any other ntervention, things seem to check out (wth
> > >> ibdiagnet/ibchecknet).  An thought?  Need any more nformatin?
> > > 
> > > Is something bouncingyour subnet or was this just what ibporttte did
> > > ? It could be if this was a coreswitch.
> > 
> > Nothing should be.  The same thing appears to happen onceevery couple days
> > -- it is very difficult to correlate wth anything.
> 
> And does it just go away ? Is some part of your subnet not accesible ?
>  
> > > Also, you may have someSMAs which have gone nonresponsive to SMPs
> > > (IB_TIMEOUs) but the links are up. I can't be surenot knowng what the
> > > exact scenario was. If you do, you will like want to chase these and do
> > > something abot them if you haven't already.
> > 
> > Hmm, what could causethat?  All my hosts are responsive whenever I check
> > (though it hasn't been during one of these stors of activity).
> 
> Are all your switches responsive ? What switches are you using ?
> 
> -- Hal
> 
> > > All the messages reltin to ACTIVE-> ACTIVE transition can be ignored.
> > > 
> > > Also, it looks likesomething i removing characters n the log.
> > 
> > Yeah, there are characters missing in the whole message.  rious.
> > 
> > Thans again,
> > --Doug

Could you guys stop sending same 10000 lines back and forth please?

-- 
MST



More information about the general mailing list