2

I'm hosting with MediaTemple on a (dv) Dedicated-Virtual 3.5 server. My site consists of a Wordpress blog, some custom PHP pages (nothing too intense), and I server 500-700 unique visitors per day.

Despite my pretty modest numbers, I suffer from regular Apache crashes on account of QoS Alerts, mostly flagged as "tcpsndbuf".

MediaTemple support -- usually tops -- has been pretty useless on this matter. I'm looking for answers as to how/why this is happening, advice on how to stop it. My website is a good portion of my livelihood, and downtime equates to lost income.

Any and all help much appreciated.

-Matt

4 Answers4

1

"tcpsndbuf" errors usually have to do with low memory. Have you looked over your error logs for clues as to what is happening?

  • No, I have not. How/where do I access my error logs? I've been told by MT reps that I should have adequate memory, and recently had the memory on my server upgraded. –  Feb 15 '10 at 22:40
  • There are several logs which may be helpful. Here's a link to a knowledge base article which gives you some good SSH commands for discovering them on the (dv). http://kb.mediatemple.net/questions/247/Common+SSH+Commands –  Feb 22 '10 at 21:41
0

Hey I'm actually having this problem on one of my sites as well. The tcpsndbuf, from what the MT rep is telling me seems to be spiking over 100%, which then makes Apache stop taking connections.

This then has to be rectified by restarting the server.

Watching the access log and tops from ssh is pretty useless, since it could really be any of page.

Also, this problems seems to be happening every month.

  • That's exactly what I'm dealing with. Have you found any causes or solutions? –  Feb 11 '10 at 02:15
0

Do the crashes happen at the same time every day or multiple times per day?

Perhaps you're running some kind of scheduled clean up job that is using up all the memory.

David
  • 3,337
  • 25
  • 20
  • Multiple times daily, on days they occur. There seems to be no pattern. –  Feb 15 '10 at 22:39
  • When Apache crashes, does that mean you have to restart Apache or the whole server to get it working again? You should look at your logs to see what happens right before the crash. By default, they're in /var/log with Apache-specific logs in /var/log/httpd but mt may have them in a different place. – David Feb 16 '10 at 00:08
  • You should double-check you're running the latest WordPress versions. Also, are those custom PHP pages you mentioned publicly accessible? Perhaps you have some sort of memory leak in them. I don't suppose you could temporarily disable them to see if it's related? – David Feb 16 '10 at 00:10
0

Band-Aid: Install Monit, to monitor your server's services and monit will restart them automatically if it fails. Look through MT's knowledgebase on how to install YUM on your server. Then search on google on "How to install Monit with YUM".

And Cache your site to hell as much as you can.

My real suggestion would be to recommend a different host, I had the same issues as you listed with MT until I took the big move to another vps provider.