I'm having a problem with a Debian 6 server where it looses network connectivity and doesn't come back (server does not respond to pings, and can't connect to any services). The problem happens sporadically. Some times it will go a month without the problem occurring, while other times it will happen within a few days of being rebooted.

When logged into the console, I confirm that the server doesn't seem to have network conectivity (can't ping for example). All services are running though (mysql, Apache, SSH, etc).

When the problem happens, the network interfaces are up (I checked ifconfig, all interfaces looked normal). I also check iptables, and verify there were no rules that would block anything. I also try running /etc/init.d/networking/restart, but that does not fix the problem. The only thing I've found that fixes the problem is rebooting.

The server is a virtual machine with a wired connection in a data centre. The company I rent the server from thinks the problem is due to high RAM usage. They said that the server was frozen due to resource exhaustion and the server stopped responding to ICMP requests, referencing this:

root@vm2:~# free -m

      total used free shared buffers cached
Mem:  8181  7816 365   0     364     6201
-/+ buffers/cache: 1250 6931
Swap: 8191    0  8191

Do you think that high RAM usage could have anything to do with this problem? I think it is something else, but I don't know what else to check.

  • 1
    That doesn't look like a system that is running out of memory. In the free output provided, you have 6GB going to the filesystem cache. If you were running out of memory it would have long since purged most everything out of the FS cache to make room for application memory requests. Not to mention it shows 0% of the swap space being used. If you were running low on memory the kernel would have decided to swap SOMETHING out to free up the physical memory. OOM may cause network service disruption but this doesn't show memory exhaustion happening. – Bratchley Mar 8 '14 at 23:22

As stated in my comment, you're definitely not running out of memory. Kind of hard to do a post mortem when you only have part of the picture (such as with your IaaS situation).

I would check /var/log/messages and dmesg to see if anything relevant was logged around the time of the outage. Chances are, though, that nothing useful will be printed out.

Depending on the type of virtualization, you may also use ethtool to see if the link is dropped. Xen paravirt won't show a link status either way but I get the following on a VMWare instance:

root@xxxxxxvld04 ~ $ ethtool eth0 | grep "Link detected"
        Link detected: yes
root@xxxxxxvld04 ~ $

IIRC KVM virtualization will also show link status. Not sure about Hyper-V but I'm going to lean towards a yes.

| improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.