On Thu, 02 Mar 2006 20:14:01 -0800, Gord wrote:
> Folks
>
>
> I am having a weirdism. I am trying to detect when a server goes away and
> comes back.
>
What a weird expression!
>
> The approach I am taking is to monitor the arp table. This in theory
> works well. The arp entry for my server goes to <incomplete> in a timely
> fashion (ok not very timelty, but acceptable). After the server
>
It is perfectly normal in an host attached onto a Ethernet network to
forget about the cached ARP entry representing another host.
> reboots, the arp entry becomes complete again. So far so good.
>
>
I assume that something process in your server reboot must do some kind
of communication that provide your server with the MAC address of the
other host. Maybe a broadcast... Maybe it's the address of a switch,
bridge or a router.
> But there is a race condition. If I call 'arp' at just the wrong time,
>
> it will hange forever. I can <ctrl>C and everything will be fine.
>
>
> I tried cat'ting /proc/net/arp Same result. The cat proc can hang
> forever.
>
>
> Anyone seen this before? I am running RHEL4 U1 .
>
>
> Thanks
>
>
> Gord
I would recommend running a protocol analyzer to check it.
TCPDump would be fine for it.
Another area of research for you is Spanning Tree.
All the bridges/switches frequently talk to each other to determine if
there is a loop in the network. Loop will kill the net...
So it may be possible that the MAC address that represent a step in the
path to the other host change as the net topology change.
Hope it help.
|
Similar Threads
Linear Mode
