From: Herbert Poetzl (herbert_at_13thfloor.at)
Date: Mon 02 Feb 2004 - 10:48:58 GMT
On Mon, Feb 02, 2004 at 11:05:32AM +0100, Gebhardt Thomas wrote:
> about once a week we run into trouble with one of our master/vservers:
> suddenly all vservers on that host seem to be offline, they do not
> reply a ping request and all tcp connections get disconnected. The
> master is fine, however. I can ssh to the master, enter one vserver
> from there and initiate a network connection from within that vserver
> to the outer world: No problem, and from that moment this specific
> vserver gets "online" again. During the failure state I have no time
> for a detailed analysis of the flaw; just trying to get the services
> running again. I was not yet able to reproduce that situation on
> a test server, however.
> Any hints what I could do?
hmm, that sounds like some router, which gets
confused by many ip addresses for one MAC ...
older cisco routers showed that behaviour, flushing
their arp caches after some overflow ...
PS: I'm not saying that it might not be vserver
related, but that was my first thought ...
> Some observations:
> * At least one time the failure happened when I caused some substantial
> load on the master server (rsync'ed a vserver to another host)
> * The vservers stay "offline" until I actively initiate a network connection
> from within the vserver (or if I reboot the vserver)
> * At least a few minutes after the failure occurred, there are nor arp
> entries for the affected vservers in the routers arp tables, even if I try
> to ping the vserver. So it seems that the master does not reply to
> arp requests for the alias interfaces.
> I'm running 2.4.24-vs1.22 (There is a HA environment with
> heartbeat and drbd on the affected host, but I dont't think that this is
> Thanks for any hint!
> Cheers, Thomas
> Vserver mailing list
Vserver mailing list