RE: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

RE: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

[Ian Pratt]

> > I guess setups with multiple active NICs are relatively 
> rare and can 
> > be dealt with using a top-level script (or configured manually).
> 
> I'd disagree with that (unless I am misunderstanding the 
> definition of active).  Every pizza box server I've seen ship 
> in the last 3 years has had dual onboard nics.  Dual nic 
> setup for server (one for production and one for 
> administration) is best practice for any reasonable size 
> server environment.

I think you're actually agreeing with me.

Having a separate dom0 interface for administration is certainly good
practice. 
However, having multiple bridges is probably relatively rare, so we can
probably get away with not having multiple network-script lines.
However, I'd prefer that we could.

> > NB: The default for 'network-script' should be to run no 
> script at all.

That's true for versions shipped with a distro, but right now, having
the network-bridge script as the default when xend starts should
actually work for most people.

It probably won't if you're using NFS root or anything else fancy, but I
still think its worth having.

The current observed flakiness by some people is disturbing and we need
to understand why. Please can someone who is seeing this problem please
dig in to it. We've run the script on lots of different machines without
seeing problems.

Thanks,
Ian

Re: RE: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

[Nate Carlson]

On Fri, 5 Aug 2005, Ian Pratt wrote:
> It probably won't if you're using NFS root or anything else fancy, but I 
> still think its worth having.

Actually, it does work on NFS root.  :)

------------------------------------------------------------------------
| nate carlson | natecars@natecarlson.com | http://www.natecarlson.com |
|       depriving some poor village of its idiot since 1981            |
------------------------------------------------------------------------

Re: RE: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

[Tom Wilkie]

When I was doing some testing a few weeks ago, I found the network 
script would knock out my NFS root for dom 0 when I started xend - this 
would in turn hang the machine.

We suspected it was when the routes / addresses got removed from eth0. 
We were using some quite new gigabit ethernet cards, ian though it might 
be because the promiscuois(?) flag was being lost.

I will do some more testing on this next week, let you know if this is 
still the case.

Tom

Nate Carlson wrote:
> On Fri, 5 Aug 2005, Ian Pratt wrote:
> 
>> It probably won't if you're using NFS root or anything else fancy, but 
>> I still think its worth having.
> 
> 
> Actually, it does work on NFS root.  :)
> 
> ------------------------------------------------------------------------
> | nate carlson | natecars@natecarlson.com | http://www.natecarlson.com |
> |       depriving some poor village of its idiot since 1981            |
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

RE: RE: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

[Ian Pratt]

> When I was doing some testing a few weeks ago, I found the 
> network script would knock out my NFS root for dom 0 when I 
> started xend - this would in turn hang the machine.

Using the network script with an NFS route is always going to be
extremely risky, unless you you copy all the binaries and libraries etc
you need over to a tmpfs ramdisk or something. 

If you're using NFS root its much safer just to have an intrd that sets
up networking as you want it first, rather than using the network-bridge
script at all.
 
Ian

Re: RE: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

[Nate Carlson]

On Fri, 5 Aug 2005, Tom Wilkie wrote:
> When I was doing some testing a few weeks ago, I found the network 
> script would knock out my NFS root for dom 0 when I started xend - this 
> would in turn hang the machine.
>
> We suspected it was when the routes / addresses got removed from eth0. 
> We were using some quite new gigabit ethernet cards, ian though it might 
> be because the promiscuois(?) flag was being lost.
>
> I will do some more testing on this next week, let you know if this is 
> still the case.

Huh, interesting - at least with 2.0, it's working just fine for me with 
the standard bridging scripts.

------------------------------------------------------------------------
| nate carlson | natecars@natecarlson.com | http://www.natecarlson.com |
|       depriving some poor village of its idiot since 1981            |
------------------------------------------------------------------------

Re: [PATCH] network -> network-bridge rename WAS: Re: RE: help with bugs

[Sean Dague]

[-- Attachment #1.1: Type: text/plain, Size: 2193 bytes --]

On Fri, Aug 05, 2005 at 03:13:33PM +0100, Ian Pratt wrote:
> > > I guess setups with multiple active NICs are relatively 
> > rare and can 
> > > be dealt with using a top-level script (or configured manually).
> > 
> > I'd disagree with that (unless I am misunderstanding the 
> > definition of active).  Every pizza box server I've seen ship 
> > in the last 3 years has had dual onboard nics.  Dual nic 
> > setup for server (one for production and one for 
> > administration) is best practice for any reasonable size 
> > server environment.
> 
> I think you're actually agreeing with me.
> 
> Having a separate dom0 interface for administration is certainly good
> practice.  However, having multiple bridges is probably relatively rare,
> so we can probably get away with not having multiple network-script lines.
> However, I'd prefer that we could.

I was thinking about this over the weekend, and it occured to me that this
might not be the case.  Today, when we are all doing testing on
xen-unstable, we're pretty much always running test domUs so have something
like:

ethX -> bridged/routed to domUs
ethY -> dom0 only, in case ethX dies

However, in many production environment you need an administrative lan to
all servers to apply software updates and the like.  So you'd actually have.

ethX -> bridged/routed production lan for domUs
ethY -> bridged/routed admin lan for domUs

Hopefully ethY would be reliable enough that we wouldn't need a 3rd NIC to
be admin for dom0 in the event of xen network stack fall over, but we might.

Not proposing any solutions as yet, but it is worth noting that this might
not be as uncommon as first thought.

	-Sean

-- 
__________________________________________________________________

Sean Dague                                       Mid-Hudson Valley
sean at dague dot net                            Linux Users Group
http://dague.net                                 http://mhvlug.org

There is no silver bullet.  Plus, werewolves make better neighbors
than zombies, and they tend to keep the vampire population down.
__________________________________________________________________

[-- Attachment #1.2: Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel