Server Reboot Frequency

[Julian Muñoz]

Hi,

I have two dedicated servers for my database and the webserver after months of uptime seems to slow down the processing speed but it all fixes with a simple reboot.

What is the suggested frequency between reboots for a dedicated server?

[thisisit3]

If you run Linux or FreeBSD then you shouldn't have a problem with 300 or more days uptime, some people run kernel versions 2.2/2.4 with several years uptime.

In other words, if you need to reboot the server then you have some kind of problem, like a memory leak (like the ones caused by spamassassin).

Try find which application causes the memory leak or see which one causes higher cpu usage over time.

[jason]

Yeah, I agree with thisisit...it sounds like a memory leak. I'd keep an eye on your memory usage for specific programs using something like the top command to see if you can spot the offenders.

--Jason

[Julian Muñoz]

The webserver only runs lighttpd and php5 running as fcgi. But even when I restart lighttpd, the webpages are really slow when uptime is like 150+ days.

I have debian and the usual tools on the standard installation. So that's really weird.

[Gwaihir]

If it's not a memory leak, then perhaps some other garbage collector doens't work as it should.

For example (if you use php sessions stored on disk): does the number of sessions and the size of each session file remain fairly constant? It could be that you're only removing old sessions on server restart. Or perhaps you allow recurring users to use the same session forever (ok in itself), but your scripts keep throwing ever more stuff in it.

Another one to check: is your number of open MySQL connections reasonably stable, or does it keep building up? If it does, you are apparently using permanent connections that don't get reused properly by your scripts.

[Ron]

Or the temp disk in mysql is now writing to hard disk
or /tmp is writing to hard disk

[Julian Muñoz]

I'll check those! Thanks for the help guys!