0

I have a server that's hosted in a colocation facility. Recently the websites it hosts have been running slowly. Sometimes a page will load immediately (like it used to always) and other times it just won't load at all, and the request will time out. Clicking the link again usually gets the page to load, almost immediately.

Nothing's changed on the server since it was installed a few years ago, so I'm having trouble working out what the problem may be.

When I ping it, it seems normal. Also, the traffic volume is not particularly high. It has plenty of free disk space. Connections over SSH seem to be fine. I've tested on different computers and different browsers, and at different sites (to eliminate my internet connection and ISP) and it's still the same issues.

How do I work out what the problem is? Any help appreciated.

Improve this question
1
  • Nothing has changed in several years? It's probably host to the most spectacular infestation of malware this side of an Iranian nuclear program.
    – womble ♦
    Aug 12, 2011 at 1:18

2 Answers 2

Reset to default
2

From the description of your symptoms, it sounds like it's either a very peaky load, or some sort of networking glitch. The latter is probably going to involve the tedious involvement of your hosting company, but if it's the former, rather than writing the whole speil here, I'll just point you to a comprehensive guide I wrote for a previous employer entitled "Hunting the Performance Wumpus", that covers everything you're ever likely to need to know about tracking down performance issues.

Improve this answer
0

You could potentially identify the culprit by taking a look at the most memory-hogging processes when the server is actually slow. A periodic check for memory status along with a memory snapshot alert via email when things are down should help you narrow things down.

If this is a Linux box (which it probably is), you could schedule a cron job that actively monitors your memory/CPU usage every 5 minutes: possibly the output of the free command and look find out when the used:total ratio is going out of hand.

When you identify a problem, your script could capture all the relevant information and send a summary email. The outputs of top, w, etc. come to mind.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .