So, I’ve been looking at the CPU usage reports that DreamHost generates for all shared users, and I found a disturbing trend: I use somewhere around 25% of the CPU minutes on the machine known as Caesar. What do I use it on? Well, on a day when I don’t access my site via SSH, I see awstats.pl running for 5424.4800 CPU seconds, attempting to generate readable statistics for me. Other big users are php.cgi at 552.6800 CPU seconds, and perl at 24.0800 seconds. That 5424.4800 CPU seconds ALONE is worth 22.608% of the CPU seconds utilized by ALL users on Caesar that day.
They’ll probably move me to a different (hopefully, more powerful) server at some point… y’know, where there might be 5 high users on the same machine, and we duke it out for access to the resources. Either that, or I can try to reduce usage. I think I might try that… perhaps running the stats every hour IS a little too much… but I actually do use that info. So who knows what I’ll do?
AWstats is a pig. It doesn’t just use CPU time when it generates reports, it uses quite a bit of CPU as when you are requesting those reports. How much traffic you are getting so you need to run AWstats once every hour? Or are you just so desperate to know about your stats?
For me, I actually copy the access log files back to my home server every day, using ssh after the log has been rotated, and then run awstats locally just to save CPU time.
I like knowing as much as I can about my statistics, and that means having fairly recently-generated stats. I used to run everything once a day, but then I don’t get any information about what’s actually happening that day… which is bad, especially on a day like yesterday, when I keep getting a bunch of bad referers - when I start getting emails from Webalizer complaining about stuff that’s going on, that’s my warning bell. I can quickly SSH into my server right away and install a block.
Your method of copying logs back to your home server and running AWStats is not feasible. It’s not even feasible for me to run AWStats on my system against logs on Caesar, because with my internet connection (56k modem), I cannot automate network-related tasks, nor can I run a linuxbox and hope to have a connection.
Anyway, I’m not too concerned… yet. Apparently, I don’t have a lot of high CPU users sharing Caesar with me.