Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I run a regular Apache + mod_wsgi + nginx website on a VDS and encountered a strange problem. The website works fine from my computer and also from our managers' one (who use a different Internet provider). However, some of our website users report to managers that they can't open it in their browser, for a long time period, while at the same moment it worked from my computer and managers'.

The question is: how should I do the diagnostics (if this bug is even my fault) and what bug should I look for if it is impossible for me (even if I wanted to) to go to the problematic user and do the diagnostics from their computer?

UPDATE: Thank you guys for your support! I am going to contact the faulty clients through our managers, do the suggested tests online with them, and then I'll provide additional info if necessary.

UPDATE #2: I managed to contact one of the problematic clients and performed the diagnostics with wireshark. It turned out that the trouble was caused by a bug in the notorious get-iana.sh script in Firehol. As a result, a certain IP address range was incorrectly identified as RESERVED. Everything worked fine... until our local ISP started using those IP addresses for their dynamic IP pool and some website users became locked out. As an aftermath, it seems that I shouldn't have used Firehol in the first place since it is no longer maintained and neither does it support IPv6. Thanks again everyone for your answers.

share|improve this question
    
How long has this been going on? Can they ping to the server? DNS ok? –  Bart De Vos Mar 31 '11 at 8:20

2 Answers 2

up vote 1 down vote accepted

I think I would check things in this order:

  1. Make sure that the failing users actually reaches the web server with wireshark or tcpdump. (My bet is they don't and you can hopefully blame someone else :)
  2. Check for spikes in load or other strange blocking in the server, e.g. do things like 'find / -type f' while the users fail. Maybe unmount any NFS volumes if there are any.
  3. Measure roundtrip in the webserver, either by turning on logging of roundtrip time (in Apache this would be LogFormat %D) or using wireshark to see if the time is consumed in the server or in the browser.
  4. Installing Firebug in a Firefox on a failing machine and look at its network traffic diagram to see what takes a long time (this is of course assuming that the previous steps have indicated that the browser actually talks to the server).

EDIT: Even if you can't go to the site of the failing browsers, maybe there is remote desktop or remote assistance to help you?

share|improve this answer

A good starting point is checking at the network level if the problem is recurring.

You could get a traceroute from the users who cannot access the site and copy and paste it to their ISP. The ISP should then be able to tell you if the trace is completing (and therefore it's a Server issue) or of it's on the network and something to do with their peering/routing.

Traceroutes can be misleading so handing it over to the ISP should get a clearer answer for you.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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