We did an upgrade to all of our OL 5.8 hosts. Ever since then we've been experiencing networking issues.
- Xen hosts running kernel-uek-2.6.39-400.17.2 kernel.
- Not able to resolve hostnames via DNS, but the dig command returns a valid entry!
- Restarting networking or rebooting the box fixes it temporarily.
Here's what's interesting.
Within /etc/resolv.conf, we poiint to the VIP of our DNS servers. The VIP is handled by our netscaler. IF I put one of the DNS server's IP into /etc/resolv.conf (and not the VIP), it works just fine. With the VIP set in /etc/resolv.conf, a ping request using a FQDN drops immediately. What's more interesting is, tcpdump does not show any traffic when attemping to ping. Dig requests shows traffic though.
Also, strace shows "network down" when attemping to ping via hostname using the VIP. All signs point to the netscaler but our OL6 hosts do not exhibit this behavior. In addition, netstat shows packet losses for all OL 5 hosts after the patches were installed.
You are describing some symptom without any technical information and it's all mixed up.
If the network goes down shortly after restart it might be due to a TCP/IP address conflict.
If a correct TCP/IP address does not resolve the problem than it could be anything from a bad device driver to routing issues, hardware or cable problems, or wrong NIC/port settings.