Web Hosting Talk : Web Hosting Main Forums : Hosting Security and Technology : Server Keeps Disconnecting

[Man or Animal]

Recently I bought a new dual Opteron server with DedicatedNow. I managed to get everything all set up and everything, and it seems like everything was going great, but the past 3 days have been hell. I mean, my server appears to go down every few hours for no reason (from what I can see). The server's load is always at a steady .2 - .4, and I haven't seen any more than that. The memory seems good too, and I have well over 1GB free all the time. All I have on this server is a plain static HTML site with a few PHP pages, a small database, and a forum with about 100-250 people on at any given time.

Just 5 minutes ago, everything was fine. I get up, leave for a minute and when I come back, I've been "kicked" out of shell saying that the software caused a connection abort, and whenever I try logging back in I get a "Network Error: connection refused" error. FTP, SSH and HTTP don't work, it's like suddenly the server decides to stop listening to any network activity. And I know it's still up: last night after it went back up, I received an email with the nightly cron details (cp update), which happened while no one could access my server. Of course, it couldn't resolve the cpanel server, so it couldn't update. BUT IT WAS STILL ONLINE!

I've been bugging DN for the past 3 days about this, and they've replaced both my network card and network cable. Still the same. They can't seem to figure this out, and I have no idea either. Anyone have any ideas?

[Patrick]

Do you have full duplex enabled on the NIC? (ifconfig)

I had a similiar problem with our new server at DN where it wouldn't properly auto negotiate the NIC to full duplex... so it had to be forced and that resolved random connection issues.

[Man or Animal]

Originally Posted by Pat H

Do you have full duplex enabled on the NIC? (ifconfig)

I had a similiar problem with our new server at DN where it wouldn't properly auto negotiate the NIC to full duplex... so it had to be forced and that resolved random connection issues.

I'm not sure how to tell... this is what I get when I run ifconfig:

eth0 Link encap:Ethernet HWaddr 00:13:46:8D:EB:88
inet addr:208.116.7.98 Bcast:x.x.x.x Mask:255.255.255.248
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:4554835 errors:0 dropped:0 overruns:0 frame:0
TX packets:5290224 errors:9 dropped:0 overruns:0 carrier:9
collisions:898345 txqueuelen:1000
RX bytes:1163701869 (1.0 GiB) TX bytes:5514366140 (5.1 GiB)
Interrupt:20 Base address:0xbc00

[Patrick]

What about using mii-tool or ethtool ?

[Man or Animal]

Ok, I used ethtool to change it to full-duplex, and I used mii-tool to check and it appears to be in full-duplex mode now. I guess we'll just have to see if this works. Thanks!

[Man or Animal]

Bah, my server just died again... does anyone have any ideas?

[Patrick]

Originally Posted by Man or Animal

Bah, my server just died again... does anyone have any ideas?

If you're running windows, do a ping www.yourserver.com -t and let it run... when you lose your connection check the pings and see if they time out.

Keep in mind that DN/FITX had issues earlier today and some routing issues afterwards however I have sent over 4000 pings without any latency or issues so it looks like everything is back to normal in regards to the network.

[Man or Animal]

Ok, I pinged my server and it got to 537 pings when I got kicked out of shell with that "software caused connection abort" error. The server was down for half an hour, went back up, and after a few hours no one can access it... again. What kind of problem does this sound like? Would it be a network issue? I tried emailing DNow's network support, but I haven't gotten a reply yet... Apparently they're not around on weekends.