How often should servers be rebooted?

[Lynne]

I would guess that the answer to this question could depend on many things, but how would someone determine just how often or if the server should be rebooted?

I run a forum that gets about 6500 unique users a day and about 800 posts a day (optimization request is here (http://www.vbulletin.com/forum/showthread.php?p=1245326#post1245326)). We started having 'server' stalls about two months ago where suddenly the cpu usage would go to 0% and nothing would happen on the server. The problem would either fix itself or I would restart apache. Then, about a month ago I hired someone to upgrade all the software (php, apache, mysql) and he also 'tuned' the server (pretty much the way George recommended in my optimization request, but a few small differences). He restarted the server, of course, and life was going along fine for two weeks until our host had some problems, and the server got rebooted. That was two weeks ago and life was going along fine until today. Now, we are back to these server stalls again.

So, I'm wondering.... is this a clue that perhaps I should reboot the server about every two weeks? Besides these server 'stalls', is there something that would clue me in to when I should reboot it?

[Indy]

This may not help much with your question, buy my server uptime has been 301 days. Now, since I upgraded from 3.5.4 direct to 3.6.4 I have noticed that it takes a wee bit more horsepower, (higher server load), to run the forums.

Typically with 3.5.4, during prime time-peak hours, my load would get as high as say .8, but usually ran around .2 to .5. Now with 3.6.4 I see it range from .3 to 2.0, but hovers around .8 to 1.1. Every once in a while things go whacko and the server load sky rockets to 20.0+ and my IO Wait is maxed 100%. This only happens when I'm working in the vB admin.

I will be moving to a more powerful server in Jan and getting away from the SATA drives moving to SCSI and uping the RAM from 2G to 4G. But, like you I was wondering the same thing. I am planning on rebooting my server tomorrow and monitoring the server load for several days afterwards to see if it indeed had any positive effect. I will re-post with what I find.

I'm also curious to see what the real wizards here think about reboots.

Dan

[kerplunk]

Windows 2000 servers: Once a week
Windows 2003 servers: Once every two weeks
Linux/UNIX servers: Only when you want to upgrade kernels

[Lynne]

301 days?!? Wow. I wish. Actually, I can't imagine. I used to reboot about every two or three months (the site is only about two years old though and has only been as 'active' as it is for about a year).

[Lynne]

Windows 2000 servers: Once a week
Windows 2003 servers: Once every two weeks
Linux/UNIX servers: Only when you want to upgrade kernels
Yep, I just got the kernel upgraded four weeks ago, and I got the software upgraded then too. The server stalls seemed to stop. But, they are back again, so I figured it must be a server reboot issue.

[jason|xoxide]

Windows 2000 servers: Once a week
Windows 2003 servers: Once every two weeks
Linux/UNIX servers: Only when you want to upgrade kernels

That's pretty much dead on although I did have a Windows 2000 Server running a VPN that ran for 421 days without a reboot. I've got a screenshot of the uptime around here somewhere...

In any case, my record on a *nix box is 677 days but that was a box on a private network. I'd never let a public box run that long without applying a kernel update.