On Tue, Aug 14, 2001 at 06:20:51PM +0200, Janice Girouard wrote:
> There are a number of patches out at the
> www.sourceforge.net/projects/bonding
> site for release 2.4.4 at:
>
> http://sourceforge.net/project/showfiles.php?group_id=24692&release_id=47592
>
>
> These patches represent work since 9/30/2000 from various individuals for
> some nice improvements in the bonding.c code. A detailed list of these
> changes are included at the bottom of this note.
>
> I was hoping that we could receive feedback on these changes to facilate
> having them accepted. If you have a moment, could you take a look at this
> work and provide input.
One big problem in my opinion is how the MII monitoring is implemented.
It uses ioctls that are marked for removal already in vger and hardcodes
data structures in an an ugly and non portable way. I think it should
use new device functions instead.
In addition it is a bit useless in my opinion. MII monitoring alone
is never enough to assure HA, because there can be lots of other reasons why
the other host can go belly up without losing the ethernet link (e.g.
a software crash). For these an higher level heartbeat is required anyways
(or at least the neighbour states in the kernel should be used which maintain
similar information at least upto L2). I don't see any interface for such
an higher level heartbeat, and if it exists the MII monitoring is not really
needed anymore.
The other stuff doesn't look too bad.
Of course there is the fundamental problem of the bounding device that
it reorders packets when more than a single interface is used in parallel
and therefore kills performance in most network protocols. If you care
about that you should use multipath routing instead (which has no
[useless] mii monitoring, but an easy to use interface for higher level
heartbeat)
-Andi (who thinks multipath routing is superior to bounding devices)
|