Re: [Xen-devel] [xen-4.0-testing test] 2045: regressions - FAIL

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Ian Campbell writes ("Re: [Xen-devel] [xen-4.0-testing test] 2045: regressions 
- FAIL"):
> This is with a PVops domU kernel? If so then I'm surprised this is a
> regression rather than a never passed. Until very recently a pvops
> kernel would not even attempt to send a gratuitous ARP after a
> migration, which could lead to 20-30s timeouts like this. (I suppose it
> might spuriously pass if a test run got very lucky?)

That's quite likely.  arp timeouts are typically a few minutes so you
have maybe a 10% chance of getting lucky.  One pass in the past and
the tests will consider it a pass.

> Even with the fix in place the gratuitous ARP behaviour is disabled by
> default so you need to enable the net.ipv4.conf.<dev>.arp_notify sysctl
> for any device you want to send the notifications. When I was testing I
> did this by adding
>         net.ipv4.conf.default.arp_notify = 1
> to /etc/sysctl.conf and that seemed to do the trick.

I think this is a bug.  I think the default should be for something to
send this gratuitous arp and the most logical answer in the PV or
PV-on-HVM case is the domU.

However: all of this doesn't apply to the test network used for these
tests.  There is a daemon on that network which sends broadcast ARP
who-has packets for every IP address at very short intervals (5s or
so) and the reply will cause the switches to update their ideas of
MAC<->port mapping.

Ian.

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