> Anurag Maskey wrote:
> >>Anurag committed the changeset that introduced this; but the changeset in
> >>question was the initial one for the hg cutover--so this change likely
> >>happened earlier than that.
> >>  
> >>    
> >That was during the early days of the Mercurial gate.  I think it was 
> >before the phase 1 gate moved from tw to hg, thus the "NO_COMMENT" 
> >comment. I'll try to dig through the teamware gate/comments and figure 
> >out where this change came from.

Thanks, Anurag.  The old sccs comments might be able to confirm what
we're guessing at right now.

On Wed, Nov 26, 2008 at 02:29:14PM +0000, Alan Maguire wrote:
> I _think_ this change was made by me. The reason,
> if I recall, was that during testing of the routing management
> changes, I discovered that the refactoring of dependencies
> changed boot order of services, and exposed
> a bug in network/physical's dependencies.
> 
> The bug was that dhcpagent - launched by network/physical's
> instances, both default and nwam -  requires loopback
> interfaces for IPC. Otherwise DHCP acquisition fails. Prior
> to the routing management changes, we were just lucky this
> didn't hit us. I addressed this as part of CR6312914.

I think you did introduce the service-wide dependency that is what we
have in ON today.  What I was wondering about was why it was split out
to be per-instance in the nwam phase 1 gate.

John's vague memory seems to confirm what I was thinking: that it went
in way back when we were thinking that nwamd would be creating lots of
additional instances of the network/physical service, and we didn't
want those instances to inherit the dependency.  But that is no longer
a concern with the current design.

If we can't come up with any other reasons for having the split, I'll
revert what's in the phase 1 gate to match ON.

-renee

Reply via email to