[ofa-general] Re: ipoib / bonding and OFED

[Bob Kossey]

Or Gerlitz wrote:
> Bob Kossey wrote:
>> Just to follow up on this, using RHEL5 and OFED 1.2 rc4, I was able
>> to do enough rudimentary testing to convince myself that IB
>> bonding was working.  I was able to use ib-bond, as well
>> as the use of the openib.conf file to enable bonding on startup,
>> including both(separately) IPOIBBOND_ENABLE and IPOIBHA_ENABLE.
>
> Thanks for the feedback. OFED 1.2 supports both options, however, I am 
> don't think that two HA solutions should be deployed at commercial 
> distributions. What is your take (bonding vs ha daemon) on the correct 
> way to move fwd?
I agree we don't need both at the same time.  I was wondering myself 
what the pros and cons
of each method were.  What are the link monitoring methods used by 
each?  The bonding
method would have the advantage of commonality with other bonded 
interfaces, and
may be simpler and more reliable than the user daemon.
>
>> One thing I was not able to do however, was to start IB bonding
>> using the standard bonding modifications to /etc/modprobe.conf
>> and /etc/sysconfig/network-scripts/ifcfg* files.  Should this be 
>> possible,
>> and are there perhaps some required settings I am missing?  I'll
>> include my file modifications and some output below.
>
> On some distributions (eg RH4 and SLES10) /sbin/ifenslave is used to 
> configure bonding through the distro /sbin/ifup scheme. The ifenslave 
> program is somehow obsoleted and is not supported under the bonding 
> modifications to work with ipoib devices.
>
> Moving forward, the way to go is using the bonding sysfs api, see the 
> files under /sys/class/net/$BOND/bonding/ and the bonding documentation.
>
> This is how the ib-bond script works and also /sbin/ifup-eth on RH5! 
> on however for OFED 1.2 we did not make it to fully examine the RH5 
> scripts to the extent i can say if you can just work with the OS 
> bonding configuration scheme not i can debug for you now why its not 
> working.
>
> Its definitely on our plan, but its P2 relative to the bonding changes 
> upstream push, let me know if you this different.
>
> Or.
>
It would be nice to be able to use the standard file modifications to 
perform IB bonding,
for consistency with how we handle other bonded interfaces.  If someone 
knows how
to do it, great, but if not, I agree it would be a lower priority 
investigation.

Thanks,
Bob