[OmniOS-discuss] omnios host goes suddenly silent on the network

[Tobias Oetiker]

Hi Eric,

Today Eric Sproul wrote:

> On Tue, Oct 29, 2013 at 10:06 AM, Tobias Oetiker <tobi at oetiker.ch> wrote:
> > ADDROBJ           TYPE     STATE        ADDR
> > lo0/v4            static   ok           127.0.0.1/8
> > fugu0/v4static    static   ok           zzz.yy.8.5/23
> > fugu1/v4static    static   ok           10.10.10.1/30
> > lo0/v6            static   ok           ::1/128
> >
> > the dropout does not coincide with a big backup job ... I am
> > running collectd on the omnios host, and it has been faithfully
> > recoding what happend on the interface while it was offline.
> >
> > The trafic stats show that packets have been coming into fugu0 but
> > only very few got sent out ... (if it happens again I will do a
> > snoop in the interface)
>
> A snoop would be my first tactic too.  Are these VNICs using VLAN tags
> or is everything untagged?

will do when I get access next ...

> >> For good measure, let's also look at `prtconf -d` to see what this igb
> >> hardware is.
> >
> >         pci8086,1d10 (pciex8086,1d10) [Intel Corporation C600/X79 series chipset PCI Express Root Port 1], instance #6
> >             pci8086,3584 (pciex8086,1521) [Intel Corporation I350 Gigabit Network Connection], instance #0
> >             pci8086,3584 (pciex8086,1521) [Intel Corporation I350 Gigabit Network Connection], instance #1
> >             pci8086,3584 (pciex8086,1521) [Intel Corporation I350 Gigabit Network Connection], instance #2
> >             pci8086,3584 (pciex8086,1521) [Intel Corporation I350 Gigabit Network Connection], instance #3
> >
> > note that the kvm hosts were able to talk via igb0 while fugu (zone0) was not.
>
> OK, so this is I350, for which support should be pretty stable (it's
> been in upstream illumos for over a year and I know Joyent deploys
> I350 heavily in their public cloud).  I don't see any open issues on
> igb or I350 that would be relevant here.

the troubling bit is that during the outage, the kvm hosts on
akami0 and nigiri0 were able to talk to the physical network just
fine, but they were not able to talk to fugu0 ...  and this is all
happening inside the crossbow switch within illumos if I
understand the concept correctly ...

cheers
tobi

-- 
Tobi Oetiker, OETIKER+PARTNER AG, Aarweg 15 CH-4600 Olten, Switzerland
http://it.oetiker.ch tobi at oetiker.ch ++41 62 775 9902 / sb: -9900