On Fri, 2010-05-14 at 12:39 -0700, Dave Stevens wrote:

Hello All,

I don't have any reason to believe CENTOS (5.4) is underlying my difficulties but it's what I've got to work with so here's the setup:

I have a 2P motherboard with 2 Opteron 2376s and two ethernet ports. The setup is fully virtualized, with Dom0, dom4 and Dom1 up and running. dom4 is attached to a dhcp service on eth0 and has no problems.

I've recently connected dom1 to eth1. eth1 is connected to a dsl modem (no router) and to the ISP's dsl signal It has a static IP address, 204.174.35.205. At intervals of a few hours the dom1 service on eth1 became inaccessible, not responding for either web or ssh access. eth0 on another provider's network remains up at this time. I can fix the availability issue by ssh'ing to the eth0 side, then ssh from there to dom1 (at 192.168.0.117) and then from the inside out ping any external address, it doesn't seem to matter which one. Then the connections get through and all is well for a while.

I have a workaround in place with this command:

watch -n 60 ping -c 1 uniserve.com

That's been running for two days and access from outside has been uninterrupted.

I don't know where to start on this one. It could be a funny router setup at the ISp's end or maybe the port (previously disused) on this board is really flaky (but then why does pinging help?). The board, fyi is a Tyan Thunder 3600m, S2932.

Ideas?

Dave

Dave,

I would check /var/log/messages and dmesg on the dom0 host and the dom1 VM and see if there are messages logged during the time frames in which you lose connectivity. Also, I know some DSL modems and/or Ethernet devices will hibernate themselves if there is less than X amount of activity during Y amount of time. Your DSL modem could be losing sync. Your default gateway's MAC address for your bridged DSL connection could be changing, although this is not very likely. You can verify the MAC address by using 'arp -an' and see if it changes over time.