Some production servers I see have uptimes of more than 500 days.

Is it a bad policy to not reboot linux servers for this long ?

Are OS structures getting fragmentated ?

Generally speaking, no there's no risk on a stable system.

If the server is stable and "the business" understands the risks of running software that may not have all the relevant security patches then there shouldn't be an issue with long uptimes.

As for 500 days..... I've seen longer

I think production servers should be rebooted at least every month.

To keep the OS and filesystems in a cleaner state.

Or is it not required ?

I think the filesystem people should change their fsck checks from every 30 reboots to a user defined value (fsck default every 180 days)

If that's what you want to do and you've the resilient infrastructure to allow you to reboot parts of it without service loss or can schedule suitable outages then go for it!!

Edit: Out of curiosity, how many servers are you managing?

Rebooting servers is based on opinion, infrastructure and time. You can find *plenty* of arguments to support what you want to do, one way or the other.

Now for something completely different: http://kernelcare.com/

There are surely open source alternatives to that,.. and it basically allows you to patch a kernel in-place without reboot, like the old Solaris boxes.

Very kewl.

-----

Addendum:

There is RISK in rebooting any system. That's why there is such an argument about whether to do it or not. The one thing most admins agree on is that there is RISK to the hardware when rebooting -- disks fail, memory goes bye-bye, fans don't spin up. Even on a 'stable' system that has not one little problem that is visible, it's a possibility that it won't boot back up for some reason. Especially after 500 days in service!

So keep that in mind when researching your rebooting cycle.

Personally mine is 'security' based. If an urgent patch comes out for the kernel, she gets a reboot. But I have spare parts, spare servers and everything I manage is in a HA configuration, so if half the hardware doesn't come back up, I'm not screwed.

1 members found this post helpful.