[nzlug] Virtualization + ethernet/ping issues.

[Michael Hutchinson]

> -----Original Message-----
> From: nzlug-bounces at linux.net.nz [mailto:nzlug-bounces at linux.net.nz]
On
> Behalf Of Steve Holdoway
>
> On Wed, 13 Feb 2008 12:19:41 +1300
> "Michael Hutchinson" <mhutchinson at manux.co.nz> wrote:
> 
> >
> > >
> > > Which version of vmware?
> > >
> > > Cheers, Rex
> >
> >
> > Sorry all for leaving this out of my post. We are using VMWare
Server
> > 1.0.3
> >
> > I believe there is an update, but are not rushing out to do this as
the
> > Virtualized windows servers are required to be live 99.99% of the
time
> > (at least, management would like that :)
> >
> > Cheers,
> > Mike
> 
> Don't forget that scheduled downtime is not included in that total! I
had
> routing issues ( debian clients losing their default routes ) when
running
> on a pre-1.0.4 vmware server installation. The host is 6.10 ubuntu -
all
> 64 bit.
> 
> Upgrade was harmless in my case - 2 debian servers and win 2003 r2
server
> clients.
> 

Hi Steve,

Something in common: our bare metal host is 64bit (AMD Athlon(tm) 64 X2
Dual Core Processor 5600+), however it runs Windows 2003 server 32bit
edition, and the guest O/S's are 32bit. The decision for 32bit was made
before I was working here.

I investigated the default route issue, and I don't think we are losing
that on the server in question, as it monitors nearly 100 sites, and
several sites on the same IP ranges as the one affected - no other sites
are reported down. Where they should be if our default route is being
lost. Keeping in mind the affected site is up and working fine, and
responds to pings from other servers.

It is possible to schedule an upgrade. It would be nice to be able to
prove that VMWARE is the issue first, however. There must be some
technical network tests I can perform to shed some more light on the
issue. Knowing that tcpdump will report that the returning packets are
fine, I'm left with stracing the ping command, which I'm off to do right
now.

Cheers,
Mike