How often do you reboot your Linux Webserver

Plenty of experience with Unix and Windows, but not Linux

[lgn1]

As everybody knows, systems even in a steady state, have a habit of self destructing, the longer they are up, due to memory leaks and other O/S related problems.

In the old VAX/VMS days, you could keep those systems up for close to a year.

With NT I rebooted, once a week to avoid problems,
Windows 2000 Server every two weeks
Unix Servers once a month.

and this appeared to work pretty well.

Im new to Linux administration, so I wonder what the recommended time between reboot, of a Red Hat linux webserver in a steady state (no patches and updates).

[Frank_Rizzo]

Even though Linux is quite robust I still tend to reboot about every month.

Not that it needs doing - just that I'd want to know that if in the event of a hang or hardware failure that it would restart again.

If I do a lot of changes tonight I'd rather not wait until a hang or hardware failure to find out if those changes screwed up the booting procedure etc.

[encyclo]

Patches and updates in Linux don't usually require reboots, unlike Windows (as you can just restart the daemon). If you don't have any particular problems then you can just restart individual daemons (eg. Apache) if there is a problem.

The only time you absolutely need to reboot is when there is a patch to the kernel - make sure the old kernel is available as an option in case the new one fails for some reason, especially if you compile your own kernel.

[StupidScript]

Yep ... I agree with encyclo. I only reboot when I update the kernel.

[wheel]

I basically don't reboot my linux webservers ever, nor do I have a reason to. If something gets locked up I just log on to the server and kill the problem process.

You've got to reboot when you do kernel patches, but I don't recall ever really doing a kernel patch. Instead every year and a half or so I generally end up upgrading my hardware. At that point I like to do a fresh install using the latest stable release of my disto, which includes the latest stable kernel.

Short answer, never need to reboot. (besides, if I rebooted regularly I'd be worried about some process not restarting :) ).

[Frank_Rizzo]

(besides, if I rebooted regularly I'd be worried about some process not restarting :) ).

That's the point I'm trying to make. You never know if all the services will start correctly unless you do a reboot.

OK, you can do a stop / start on a particular service but that is not a 100% check.

[martin]

I remember that once after about 3 months uptime on my home gw box, it turned out after I restarted it a few services that I had "temporarily" started before didn't start up. Just had to add them to the boot start up sequence.

It's easier to just check once in a while the output of rc-update show or what's relevant on your distro.

[wheel]

>>You never know if all the services will start correctly unless you do a reboot.

Well since I never reboot, it's never a problem. And there's never any other reason to restart the machine. On some odd chance that something catastrophic happened and I ran into this, I'd fix it at the time instead of going to look for it. I also run my servers relatively lean so I tend to have few services running.

I suppose there's a case to reboot once in a while - if the service startup is a concern. It just seems like work to me, for no benefit. I've been managing for 5-6 years this way without any problem :).

[kullervo]

Every time I do hardware maintenance witch occurs about every 3 months. I have a memory leak in courier so I restart it every night.