Strange Apache2 freezes?

Question

Apache HTTP Server

I ran into this problem: quite a standard VPS, Apache2 MPM prefork, Nginx for static, ~50 not very complex sites, about 10 gigabytes of daily traffic, 600,000 files. The problem is this: periodically the generation time rises to 3-10 seconds. Those. we walk around the site - pages are generated in 200 ms, and then rrraz and several times are given from 3 to 10 seconds.
In the error logs - cleanliness, in top - everything is close to zero, CPU - 0-20%, free RAM is more than a gig.
I have other similarly configured VPS with a similar load, they do not have such a problem, and on the “problem” VPS, up to a certain point, everything was fine and nothing changed in the configuration.
Perhaps someone has encountered a similar problem, or at least tell me which way to dig.

Answer 1

HDD?

Answer 2

See what WA is in the top. If this parameter is large during your suspension. Some of your neighbors are raping dicks.

Answer 3

if it's Virtuozzo linux, then cat /proc/user_beancounters and check which parameter gives any values ​​in the failcnt column. Usually bottlenecks are:
kmemsize
numothersock
dgramrcvbuf
if any values ​​are non-zero in this column, you can pull the hoster to reconfigure the VPS.

Answer 4

No, these are just limits: soft and hard. In this case, perhaps something slows down in the scripts / databases themselves. Set up the output of slow queries on the mysql server, run iostat with a log entry, you can also run something like ntop with a log entry. It doesn't hurt to check the cron jobs if the brakes are observed at a certain time (for example, Apache or Njinx logs can be rotated at such times, files are indexed on disk, etc., etc.).

Answer 5

Try the atop utility , it helps a lot when you don't quite understand what the problem is. Pay attention to the use of the screw and the network during the moments of hanging. I also highly recommend installing Nagios . He will tell you a lot of interesting things.