| Netrork dropping after a period of time in guest OS [message #49093] |
Tue, 19 March 2013 22:48 |
kubla
Messages: 1
Registered: March 2013
Location: United Kingdom
|
Junior Member |
|
|
Hi Folks,
This is a bit of a long shot, but here goes: I have 2.6.18-274.3.1.el5.028stab094.3.owl1 x86_64 installed on a Dell Poweredge 1950 with a variety of Ubuntu guests. The on-board Broadcom NIC's have failed on this particular machine, so it has a single Intel PRO/1000 NIC which was configured to run on one IP range. This config all works fine.
I had reason to move a couple of VM's from an identically configured server which needs maintenance; however they use a different IP range. I configured a virtual interface eth0:1 on the host with a spare address in this range & successfully brought the new guests up, however they lost network connectivity after about 3 hours. I restarted the VM's & they worked again for about another 3 hours before losing network connectivity again. I haven't got precise timings yet, but this smacks of some kind of timeout. The VM's on the physical interface eth0 are unaffected, it's only the 2 on the virtual interface eth0:1 which go AWOL. The guest OS can be eliminated since there are VM's on the physical interface using the same venet config:
/proc/sys/net/ipv4/conf/venet0/proxy_arp = 1
/proc/sys/net/ipv4/conf/venet0/forwarding = 1
I have a kludge in place which pings the VM's in question once a minute & issues a vzctl stop, sleeps for a second then issues a vzctl start on the affected containers if the VM doesn't respond to the ping. Obviously far from ideal, but it works.
My Google-fu appears to have deserted me which means I'm probably looking at an edge case. My questions are as follows:
1) Has anyone else experienced anything like this?
2) Is anyone aware of any ethernet timeout things I can tweak?
3) What can I use to debug this issue further?
Kind regards,
Kubla
|
|
|
|
OpenVZ Forum
Members
Search
Help
Register
Login
Home
