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

I measured load times for a wordpress 2.9.2 install on apache 2.2.3 and I was intrigued by the long periods between connect and first byte for the css and image files.

Load Average is 0.0, 0.0, 0.0 and there is 150MB free RAM on the VPS. There are a few other VPSes on this box, with even more free RAM and underutilized CPU, that have the same issue. Sample Pingdom results are at http://imagebin.ca/img/6UaiOU.png

How do I gain insight into the possible causes of this problem and how would I troubleshoot it?

Thanks

share|improve this question
add comment

2 Answers 2

up vote 1 down vote accepted

You should diagram, isolate and observe all possible paths for latency. It appears that all I/O-related requests (files) are taking much longer than the dynamic content. Use iostat. Try to compare with other local I/O operations and benchmarking tools (dd, bonnie) and see if it looks like "normal" behavior for this system or something specific to Apache.

share|improve this answer
    
i ran IOstat, on the hardware node (since the problem occurs on all VPSes in this HW node), disk %util over 5 minutes is at most times under 5%, goes to ~20% and once went to 50%, but it is clear the disk subsystem has plenty of capacity. still intrigued. –  Gaia May 16 '10 at 23:18
    
I concluded that there is nothing wrong with my server. Other measurement tools show very responsive times: Yslow, site-perf.com and uptrends.com. I'm going to write this off as Pingdom/Octagate simply "don't like" my server or its network. I COULD create another question to find out why pingdom does this all sites on all VPSes on my server. I might get to that some day. –  Gaia May 17 '10 at 13:01
    
If you'd like to test your system with Circonus (circonus.com), I can hook you up with a trial account. Shoot me a tweet (same contact) if you're interested. –  obfuscurity May 17 '10 at 18:57
add comment

can it be that somewhere you check rev-dns for the ip connection comes from? make sure neither apache or mysql do no try to do it:

maybe there are some other checks - try running tcpdump on loopback and wan interface while running single http query and see what happens.

share|improve this answer
    
probably not mysql related because images and js do not generate a DB hit, but on the apache side it is set to off. –  Gaia May 15 '10 at 18:21
    
Also skip-name-resolve is only for remote MySQL connections. Most likely they are using a local MySQL server with a socket connection. –  reflexiv 12 hours ago
add comment

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.