[Users] Re: ARP entries being ignored (once again)

[Vitaliy Gusev]

On 22 October 2008 19:05:40 Aleksandar Ivanisevic wrote:
> Vitaliy Gusev <vgusev at openvz.org> writes:
> 
> > On 22 October 2008 12:25:02 Aleksandar Ivanisevic wrote:
> >> 
> >> Hi,
> >> 
> >> I am having similar problems as described in this thread
> >> 
> >> http://www.mail-archive.com/users@openvz.org/msg01546.html
> >> 
> >> basically proxy ARP entries for containers disappear from the arp
> >> cache after some time, but, unfortunately, there is no solution to
> >> this other than restarting, at least according to that thread.
> >
> > Are you sure that device eth0 didn't go down/up ? Down/up commands
> > clean arp cache and rules, so arp entry for VE will lost. This can
> > happens when link of the device looses carrier.
> 
> Indeed, network went down for some reason. Actually it went up, but I
> can't see link going down in the syslog.
> 
> Oct 22 05:21:45 ovz1 kernel: eth0: link up, 100Mbps, full-duplex, lpa 0x05E1

It is a driver message and driver merely didn't print "link down".

For RedHat you can create /sbin/ifup-local script which will restore arp table.
This script is called when interface goes up.


> 
> this is openvz running in a xen hvm which was paused and resumed for
> backup, most probably this is whats causing it.
> 
> Thanks for the hint.
> 
> 



-- 
Thank,
Vitaliy Gusev