[Ocfs2-users] Cluster setup

Ulf Zimmermann ulf at atc-onlane.com
Fri Oct 12 08:01:35 PDT 2007


You have Oracle people telling us not to use bonding.

> -----Original Message-----
> From: Sunil Mushran [mailto:Sunil.Mushran at oracle.com]
> Sent: Thursday, October 11, 2007 15:28
> To: Ulf Zimmermann
> Cc: Randy Ramsdell; ocfs2-users at oss.oracle.com
> Subject: Re: [Ocfs2-users] Cluster setup
> 
> How is this a fs problem?
> 
> Ulf Zimmermann wrote:
> > We don't and when we were investigating why we had on the ProCurve
> > 4108gl reassembly problems, we were specific asked if we are doing
> > bonding or VLAN tagging (neither we were doing). Just looks like the
> > ProCurve are loosing packets without telling so. We switched in
Cisco
> > 2960G-48 with Jumbo Frames now and haven't had any reassembly
timeouts
> > since then. Global Cache timeout has gone down significant. Each
> > Interconnect for Oracle 10G has its own Cisco 2960G-48 now.
> >
> >
> >> -----Original Message-----
> >> From: Sunil Mushran [mailto:Sunil.Mushran at oracle.com]
> >> Sent: Thursday, October 11, 2007 15:13
> >> To: Ulf Zimmermann
> >> Cc: Randy Ramsdell; ocfs2-users at oss.oracle.com
> >> Subject: Re: [Ocfs2-users] Cluster setup
> >>
> >> Use network bonding.
> >>
> >> Ulf Zimmermann wrote:
> >>
> >>>> -----Original Message-----
> >>>> From: ocfs2-users-bounces at oss.oracle.com [mailto:ocfs2-users-
> >>>> bounces at oss.oracle.com] On Behalf Of Alexei_Roudnev
> >>>> Sent: Thursday, October 11, 2007 11:10
> >>>> To: Sunil Mushran; Randy Ramsdell
> >>>> Cc: ocfs2-users at oss.oracle.com
> >>>> Subject: Re: [Ocfs2-users] Cluster setup
> >>>>
> >>>> I explained you:
> >>>> 1 - single heartbeat interface IS A BUG for me.
> >>>>
> >>>>
> >>> I haven't really followed the whole discussion but that point
above
> >>>
> > did
> >
> >>> just come to my mind a few days ago when we replaced our HP
ProCurve
> >>> 4108gl used for 3 separate Interconnects on 10g, where only 1 also
> >>> carries the OCFS2 heartbeat. So if that switch dies, OCFS2 will go
> >>>
> > down
> >
> >>> while Oracle 10g could survive (if OCFS2 wouldn't die).
> >>>
> >>> I have to agree that is a bad design at this point. Heartbeat
should
> >>> also be on at least 2 links for OCFS2.
> >>>
> >>> Ulf.
> >>>
> >>>
> >>>



More information about the Ocfs2-users mailing list