Unix server should be rebooted only when it is necessary.
That is not often, typically only when upgrading the operating system or adding/removing hardware. Some of these tasks can also be done when the system is up.
In general, if nothing is changed in the system, there is no need to reboot at all. Many services can be stopped and restarted individually to make changes take effect.
The pros and cons of a complete reboot depend on the services that are handled by the server in question.
I never reboot unless there's a good reason to. It could be argued that an annual reboot is a good idea to check that all your hard drives are still healthy and can spin up, especially with older hardware, but I prefer to wing it. ;-) Annihilannic.
We never reboot a system just for the sake of rebooting. We have a couple systems that have gone 3 years or more before being rebooted. One system had reached about 5 years of uptime, but had to be rebooted for some Y2k remediation stuff. Generally OS patches and application upgrades are the only reason we reboot a system.
Now our NT systems get rebooted all the time to get back memory lost from memory leaky programs and flaky apps. We have some NT systems that have a nightly reboot for no other reason than to enable them to survive the next day!
I am going to take the opposing view, twice a year we have our UPS serviced, I prefer NOT to run my Mission Critical applications with no UPS, so I down my servers about every 180 days.
Just a bit of philosophy, but for me reiable is integrity, not uptime I tried to remain child-like, all I acheived was childish.
I prefer taking my systems down at least monthly, maybe not to a full power off but at least to reboot stage.
While they are at reboot is the time to test the UPS.
I will mention that I had a Xwnix system that was rebooted once during a 10 year period for a serial port problem. Ed Fair
Any advice I give is my best judgement based on my interpretation of the facts you supply. Help increase my knowledge by providing some feedback, good or bad, on any advice I have given.
To add my two-pennorth, I've long been of the view that rebooting on a regular basis can only be a risk to devices and boards within the system, placing uneccessary strain on them in terms of power and mechanical performance. As someone else said 'if it works don't fix it'!
we have a 10 week rolling backup using ufsdump ... after 10 weeks we do a level 0 dump, which for safeties sake we do in single user mode ... because of various daemons that might still be running (samba clients have a tendancy to stay around sometimes even if inetd is killed off) so i 'reboot -- -s', wasteful, i know but it works.
Although at one time I didn't reboot our Unix Servers except when necessary, we now schedule a weekly reboot every Sunday night. We had an issue with Informix not releasing memory running us low on memory at times and this cleared up the problem. We've been doing this for about a year or so and we've experienced no negative impact. Actually my confidence in the machines has grown now that we do this. Apparently most of those responding have not suffered power losses or failures that resulted in a reboot and then you find out that the machine won't reboot due to some long ago installed software, upgrade, or configuration change that apparently didn't seem to have a problem but may only be noticed during a reboot.
yeah, i had one of them ... someone had screwed up my /etc/hosts file ... or rather had unlinked it from the /etc/inet/hosts ... but since we run nis there was no trouble ... when i found it out i fixed the hosts files, by wiping them ...
as i said; nis made sure this wasn't a problem.
however at a reboot 4 weeks later i realised my error when 1 out of 3 interfaces failed to come up, causing a router to not route, and a print server to not allow lpd printing on 2 out of 3 interfaces ...
I like to wrap my UNIX server in a tastefule evening gown, hang natural pearls from it's drive bays and put high heel shoes on its feet... Then I spin it around and float it through the air like the dancers in Swan Lake. It lands lightly on the pavement outside spreading itself lovingly over the road. I then whisper in it's ear "Now, think about this when next you want to KTRAP"
Now thats graceful downtime...
although not really a 'reboot'.....
But I know you'd all love to do this.
(The pavement outside is already littered with the remnants of several NT4/WIN2K boxes... they bounce too )
RealSharper
"Crash, I'll crash you in a minute if your not careful"
Compliments to you all. This stunk of a troll if you ask me and none of you bit!
I had a machine chugging along for 496 days before it had to go live in a different rack. It had only been rebooted a handful of times in it's entire life anyway (y2k patches, a dire ups failure, the direly failed ups's replacement, maybe a time or two more, so the box was basically on until retirement, ~6 years. --
Andy
Just two reasons for rebooting against this general oposition (sure there are many more):
1.- Everytime you add, configure, etc... new software a reboot is recommendend if you can do it. But why? I think it's saffer to reboot when you are installing the software to see if the system goes up correctly and everything works fine, than get a surprise many time later when you are forced to reboot for any other reason and see that something goes wrong (probably you wont remember too much about the changes made, another person can be in front of the machine when rebooting, etc...)
2.- If you see that your machine generates many zombie processes, you'll have to reboot to kill them (is there any other way to eliminate them?) (and of course correct the problem generating thos zombies). Other way you can reach the max number of running processes.
BUT IN GENERAL, THE RULE IS TO REBOOT ONLY WHEN YOU CONSIDER THAT IS A NEED!
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.