Server connectivity Issue. Help needed!

[guessswh0]

Hello,

I am having a fairly big issue at the moment and hope that you all can help. I'm trying to introduce people at work to Ubuntu, and running it as a server. I have it completely installed on a server box we have.

The only services installed on this box are OpenSSH, Apache, PHP, and MySql, just the very basic LAMP setup. I have statically assigned it an IP address.

The server worked fine for about 4 days, and now it seems to randomly lose network connectivity. By that, I mean, I cannot hit it from my workstation. The only thing it is being used for at the moment is to host a wiki, so this is obviously an issue. I can walk to the server, and apt-get (to test connectivity) and it works perfectly. When I am at the server console, I have no issues. Additionally, when the connectivity goes down, I walk to the server, run an apt-get, and connectivity goes back up. You'll then be able to hit the server for roughly 30 minutes, and then it goes back down, and requires you to walk to the server again.

I've checked syslog, and there is nothing there. The only entry is PHP's hourly job that purges active sessions. I've tried commenting out the job to see if that would work, but it still loses connectivity.

In case it helps at all, this is the syslog entry that is repeated every hour:

Oct 4 17:17:01 [SERVERNAME REDACTED] CRON[2272]: (root) CMD ( cd / && run-parts --report /etc/cron.hourly)
Oct 4 17:39:01[SERVERNAME REDACTED] CRON[2277]: (root) CMD ( [ -x /usr/lib/php5/maxlifetime ] && [ -d /var/lib/php5 ] && find /var/lib/php5/ -type f -cmin +$(/usr/lib/php5/maxlifetime) -print0 | xargs -n 200 -r -0 rm)
Oct 4 18:09:01 [SERVERNAME REDACTED] CRON[2290]: (root) CMD ( [ -x /usr/lib/php5/maxlifetime ] && [ -d /var/lib/php5 ] && find /var/lib/php5/ -type f -cmin +$(/usr/lib/php5/maxlifetime) -print0 | xargs -n 200 -r -0 rm)

I'm afraid people are going to lose interest in using linux as a viable option for a server OS. I've never had this problem on any of the other Ubuntu boxes that I've setup as a server. Can anyone provide some insight?

Thanks for the help.

[CharlesA]

If it started to randomly start losing connection and there were no changes, I'd look at the NIC to make sure it's not having a problem.

Does the same thing happen if you do a clean install of server with a static ip by default?

[guessswh0]

the only change that was amde was about a week or so ago with the latest patch for the x64 kernel. But in syslog, I don't see any kernel panic messages or anything like that.

I don't know if this helps, but I have seen a random entry in syslog saying:

init: tty1 main process ended, respawning

But that's the only other entry.

If it was a NIC issue, wouldn't it not only prevent incoming connections, but also outgoing connections? At the moment, it seems to randomly block all incoming connections 30 minutes in, but allows outgoing connections (when testing from the console). I've also double-checked, and there are no ranges being blocked in /etc/hosts.deny

It will just randomly cut out in the middle of a ssh connection and I am unable to reconnect unless I go down to the server and have it run an apt-get (anything that causes it to connect to an outside network.


EDIT: As I see I never posted this before, the server is running 10.04.1


EDIT2: It looks like I am not the only one having this issue... - http://serverfault.com/questions/142...u-server-10-04

[CharlesA]

Any firewall rules in place?

If the nic was having problems, it shouldn't send any traffic, but I've seen cases where network communication was intermittent.

[guessswh0]

I have no firewall rules in place on this. Reason being is that this is already on a protected internal network for internal use only, and also because I wanted to make sure that nothing would be blocked (when I installed it about 2 weeks ago, and just haven't implemented iptables yet).

I just created a script that performs an apt-get when called (as that is the command which seems to restore connectivity, and made a cron job that will run it every 3 minutes. Hopefully that'll keep the server online until I can find an actual fix.

[CharlesA]

Double post.

[CharlesA]

Nice workaround.

The strange thing is that that fixes the communication issue.

EDIT: Saw yer edit, I guess it's a good thing that you aren't alone.

Too bad there isn't any fix for it outside of using the network to get communication back up and running.

Instead of using apt-get maybe just ping the gateway for 10 count and throw that in a script?