Monday, January 31, 2022

Re: apu2e4 intermittent network freeze

On Mon, Jan 31, 2022 at 6:03 AM Hrvoje Popovski <hrvoje@srce.hr> wrote:
>
> On 31.1.2022. 13:44, Łukasz Moskała wrote:
> > W dniu 31.01.2022 o 02:44, Amarendra Godbole pisze:
> >> My home network has a PC Engines apu2e4 running OpenBSD 7.0, acting as
> >> a firewall/router, dhcp server, and DNS server. A Ruckus wifi AP
> >> receives a fixed DHCP address from apu2e4. All devices connect to the
> >> AP, and receive IP address in the same subnet. apu2e4 has em0, em1 and
> >> em2, of which em0 is uplink from Comcast, and em1 and em2 are fixed to
> >> 192.168.10.1 and .2 respectively. I have dhcpd and unbound listening
> >> on both em1 and em2.
> >>
> >> Normally my laptop that receives an IP of 192.168.10.105 is able to
> >> ping the ap2e4 at 192.168.10.1 (and even ssh into it). Today I lost
> >> that connectivity first, and ping stopped working. A restart of
> >> network on apu2e4 got it working again. The problem kept repeating
> >> every few minutes (maybe 5 or so?), till I restarted network on the
> >> apu2e4/OpenBSD host.
> >>
> >> What changed today? In the morning, I applied the last two patches 009
> >> (expat) and 010 (vmm). So I uninstalled those, but as guessed, the
> >> problem did not go away. So now I switched to the other channel (em1),
> >> and the connectivity has been stable so far.
> >>
> >> I am completely in the dark here and do not have a clue as to what may
> >> have happened - something to do with networking, and possibly an
> >> ethernet channel going bad on apu2e4 since the second one works? Can
> >> anyone provide a few pointers on where I should start looking?
> >>
> >> Thanks in advance. dmesg attached.
> >>
> >> -Amarendra
> >
> > So, you have em1 with 192.168.10.1/24 and em2 with 192.168.10.2/24?
> >
> > Having two interfaces in the same subnet is a bad idea (unless they are
> > in seperate routing domains)
> >
> > I think that what you want to do is:
> > - create bridge0
>
> you mean veb(4)? right? :)
>
> > - move 192.168.10.1/24 address to bridge0
>
> and vport(4) :)
>
>
>
> > - remove IP address from em1 and em2
> > - attach em1 and em2 as bridge0 members
> > - make dhcpd, unbound and whatever listen on bridge0
> >
> > Alternatively, change em2 IP address to be in other subnet than em1, for
> > example 192.168.20.1/24
[...]

Thanks for your response(s). A few releases ago I did have a bridge,
but realized it causes an overall throughput drop rather than using
individual interfaces directly. I should have clarified -- even though
both interfaces are on the same subnet, only one is connected at any
given time, until yesterday, when I started seeing the issue on em1.

Let me give a try to veb(4) and vport(4).

-Amarendra

No comments:

Post a Comment