It is an age-old question among system administrators.  When do you need to reboot your server?  Unlike desktop computers or laptops, dedicated serversare “dedicated”, meaning they are expected to be available 24 hours a day and 7 days a week.  Most consumers and users are tolerant of the occasional service interruption, but with plenty of competing server hosts on the market, most expect 99% uptime.

With such expectations, it is important to limit reboots to only the most necessary situations.  With old intranet mainframes and application servers, it was common for Unix system administrators to reboot every month, just as a general maintenance procedure.  With faster servers and boot times, it may even be tempting to continue the practice, but it is probably not necessary.

With modern servers, whether they run Linux, Unix, or Windows, rebooting is really only necessary when the system kernel receives an update or some hardware maintenance or upgrades are required.  For this reason, it is important to use an operating system that is not updating the kernel every other week.  That is why certain Linux distributions are designed for servers and others are not.

Many sysadmins will boast about their servers having uptimes of 100, 200, or even over 365 days.  While this may be admirable from a stability perspective, it does raise questions about security and whether those servers have received the proper updates.  The best course of action in most situations it to take the middle ground.  Do not reboot for the smallest things, like web server runaway processes, but you should also be weary of never rebooting, since some critical security updates will require it.

By: Tavis J. Hampton