Re: NetBSD 5.1.2 under Xen 4 HVM

[Chavdar Ivanov <ci4ic4%gmail.com@localhost>]

On 9 March 2012 09:40, Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote:
> On Fri, Mar 09, 2012 at 07:55:23PM +1100, Malcolm Herbert wrote:
>> I have a VPS provider that I've just built a NetBSD VM with that is
>> apparently running as a Xen 4 domU in an HVM configuration however I'm
>> seeing a large number of
>>
>>   re0: watchdog timeout
>>
>> messages at regular intervals (seems to be every 8sec or so) from the
>> kernel (along with associated network hiccup of around 1-2sec) ...
>>
>> What is causing these, and how can I fix the problem? I don't know the
>> exact version of Xen but can find out if that's necessary.
>
> The re emulation in qemu (at last the version of qemu used by Xen) is
> buggy. You can workaround it by using the rtk driver instead:
> either use 'boot -c' and then:
> disable re
> quit
> or rebuild a kernel without the re driver (the kernel will use
> rtk instead)
>
> --
> Manuel Bouyer <bouyer%antioche.eu.org@localhost>
>     NetBSD: 26 ans d'experience feront toujours la difference
> --

I am repeating myself as of a few weeks ago - the best as far as
NetBSD under XenServer is concerned is to run it on PV mode with
XEN3_DOMU kernel. Of course you need to copy your kernel to
/boot/guest directory on the server and change the OS-boot-parameters
to "boot_verbose=1,vfs.root.mountfrom=ffs:/dev/xbd0s1a,kern.hz=100",
have the PV-kernel parameter set to the kernel (  xe vm-param-get
uuid=db74941f-2ccc-9c01-ecb2-bc2e26ffce19 param-name=PV-kernel returns
for me /boot/guest/netxen ).

Iperf gives me:

 xci@loan8 ~ % iperf -c support7
------------------------------------------------------------
Client connecting to support7, TCP port 5001
TCP window size: 32.5 KByte (default)
------------------------------------------------------------
[  3] local 10.1.32.165 port 60653 connected with 10.1.32.7 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   102 MBytes  85.5 Mbits/sec
<=============== fully VT 6.99.3 with rtk0, amd64
xci@loan8 ~ % iperf -c support8
------------------------------------------------------------
Client connecting to support8, TCP port 5001
TCP window size: 32.5 KByte (default)
------------------------------------------------------------
[  3] local 10.1.32.165 port 60650 connected with 10.1.32.8 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   660 MBytes   553 Mbits/sec
<================ FreeBSD 9.0, amd64, XENHVM kernel
iperf -c support8  0.06s user 6.98s system 69% cpu 10.102 total
xci@loan8 ~ % iperf -c support6
------------------------------------------------------------
Client connecting to support6, TCP port 5001
TCP window size: 32.5 KByte (default)
------------------------------------------------------------
[  3] local 10.1.32.165 port 60646 connected with 10.1.32.6 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   561 MBytes   470 Mbits/sec
<================= NetBSD 6.99.3 amd64, MYDOMU kernel
iperf -c support6  0.08s user 7.41s system 74% cpu 10.112 total

(the last is just XEN_DOMU with MULTIPROCESSOR; now that Manuel has
enabled multiprocessor support in XEN_DOMU by default in -current, I
will not have to build it myself).

I switched my NetBSD build machine to the latter - 4 CPU DOMU with 2GB
memory - which has been very stable the last couple of weeks (there
were a few hangs before that).

Chavdar

-- 
----