Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Restarting a server each 90-100 days really helps?
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Restarting a server each 90-100 days really helps?

dedicadosdedicados Member
edited February 2013 in General

Hello, i want to know if is really important to be always up, what i mean.

when my servers have 90-100 days of uptime, i restart them on the night, no downtime, maybe 1-2min only,

why i do it? maybe just to refresh all the services, and clear maybe some cache, is just a doubt, it really helps? or im loosing time?

thanks guys.

«1

Comments

  • jhjh Member

    You shouldn't have to do this

  • shovenoseshovenose Member, Host Rep

    Shouldn't have to do this. Waste of time and pointless.

  • unless you're specifically seeing decreased performance, there's really no need, especially if you're running headless *nix. If you are seeing decreased performance after a while, you'll want to fix whatever is leaking memory/clogging disk I/O instead of rebooting every X days

  • I restart all my servers on a daily basis. I do not enjoy my penis growing just by showing off my uptime output.

  • If you're not running ECC RAM, it's usually a good idea to reboot normal desktop RAM every once in a while.

  • DamianDamian Member
    edited February 2013

    We reboot ours every ~100-150 days for kernel upgrades, and we send out notices days in advance. We did not get on the ksplice bandwagon in time :(

    If you're not doing it for kernel upgrades, then there's no reason to do so.

  • @Damian: As far as I know, ksplice itself is free software, so you should be able to roll your own patches and deploy them using ksplice. How? Well, you're on your own on that part.

    As for rebooting... should only be necessary if you're running Windows servers.

  • raindog308raindog308 Administrator, Veteran

    Wait...we can make our penises grow by showing off uptime output?

  • raindog308raindog308 Administrator, Veteran

    @NickM said: As far as I know, ksplice itself is free software,

    I thought it wasn't...

    http://www.ksplice.com/

  • jarjar Patron Provider, Top Host, Veteran
    edited February 2013

    @raindog308 said: Wait...we can make our penises grow by showing off uptime output?

    Why do you think @miTgiB never brags about himself around here? No need to over compensate.

    That was my creepy comment for today. Thanks for listening.

  • thanks guys, then, im going to keep servers as it.

  • Why would you reboot if it is not needed :(. One thing I always worry before rebooting a node, "What if this baby never wakes up :0".

  • @raindog308, you just need the source of the currently running kernel, and a diff of the changes, and ksplice will work it's magic, apparently. It's available in the debian repos, and I'm sure there are builds of it available for EL clones too.

  • I still like to reboot some servers some times - be it just to have the 180+ days fsck run at a predictable time. You never know when the shit will hit the fan and then you will find yourself scrambling to fix things in a hurry, instead of being prepared for it.
    Of course for servers hosting customers the situation is different as some customers are very sensitive. Fortunately there are enough critical kernel updates popping up from time to time to justify a scheduled reboot 1-2 times a year.

  • On Windows, a reboot is a best practice because lots of applications have horrible memory management; the "enterprise" applications are one of the worst offenders.
    On Unix, a reboot will likely make things worse. If something has gone wrong and you did not notice before, the reboot could end in a unexpected state. This is expecially true on a physical server. Your hoped 1 minute downtime could become a nasty outage. I periodically reboot my Linux servers, to be sure that all startup scripts are properly up to date; I catched many errors this way. The reboot is scheduled on idle times, so i can fix the issues with ease.

  • jon617jon617 Veteran
    edited February 2013

    I agree. Windows often performs better after reboots. Linux doesn't, in fact, can perform worse initially as apps and the OS re-cache your frequently used things into RAM.

    However, one thing I can add. If your machine is out of all RAM (free+buffers+cache) and your running apps (like a database) have already put a lot of stuff into swap, there is no way that I know of to move the stuff from swap RAM into real RAM. Only way I know of is to restart the app that put the stuff in swap, or just reboot the whole machine (which does the same thing).

    So my question is... are you running Linux and are you running out of all RAM (the sum of free+buffers+cache from /usr/bin/free)? If so, then yes, rebooting will help your performance by clearing out the swap, but will not fix your problem.

  • @jon617 said: So my question is... are you running Linux and are you running out of all RAM (the sum of free+buffers+cache from /usr/bin/free)? If so, then yes, rebooting will help your performance by clearing out the swap, but will not fix your problem.

    negative,

    Uptime 116 days 8 hours 22 minutes
    Load Averages 0.41 0.34 0.30

    free

             total       used       free     shared    buffers     cached
    

    Mem: 16074300 14087504 1986796 0 1309672 10804456
    -/+ buffers/cache: 1973376 14100924
    Swap: 4194296 5300 4188996

  • I don't even reboot my windows laptop that often

  • @jon617 said: So my question is... are you running Linux and are you running out of all RAM (the sum of free+buffers+cache from /usr/bin/free)? If so, then yes, rebooting will help your performance by clearing out the swap, but will not fix your problem.

    To fix the problem, you reconfigure the problematic service(s), so that they are restricted in how much memory they can use.

  • jon617jon617 Veteran
    edited February 2013

    @sleddog said: To fix the problem, you reconfigure the problematic service(s), so that they are restricted in how much memory they can use.

    Right, I'm just saying, say you have a mysql database configured to use 14GB RAM on a 16GB machine, then a new process (call it, process Foo) comes along and uses 4GB RAM for a while, mysql starts putting some of its 14GB allowance into swap, then you stop process Foo, freeing up 4GB RAM but mysql already stored a bunch of stuff into swap and it won't go back into physical RAM right away. It's definitely a user error in that case, I blame the person who started process Foo. Add Java processes and other running things that may have also put themselves into swap, I don't know what processes are in swap, and it becomes one of the only cases where I would want to reboot a Linux machine to get my performance back.

    If you never swap, I don't see a reason to reboot.

  • @vedran said: I don't even reboot my windows laptop that often

    Woa, what kind of laptop is it to run 3 months nonstop?
    Mine always freezes for running 2-3 days continuously.

  • 10:44PM up 1655 days, 12:14, 2 users, load averages: 0.03, 0.04, 0.00
    USER TTY FROM LOGIN@ IDLE WHAT

  • sleddogsleddog Member
    edited February 2013

    @jon617 said: Right, I'm just saying, say you have a mysql database configured to use 14GB RAM on a 16GB machine, then a new process (call it, process Foo) comes along and uses 4GB RAM for a while, mysql starts putting some of its 14GB allowance into swap,

    In this case you've over-allocated memory to MySQL. Re-configure it to use less memory. If that's insufficient for performance then you need a dedicated MySQL box where no other processes will compete with MySQL for memory. At 14GB memory usage I'd say that's the case ;)

  • http://p.voltaire.sh/4

    you dont need to reboot to clear swap

  • @fly said: you dont need to reboot to clear swap

    Sweet! Thanks. I love learning new tricks.

  • @DalComp said: Mine always freezes for running 2-3 days continuously.

    Time for a reload. :P

  • shovenoseshovenose Member, Host Rep

    Or a new laptop.

  • ztecztec Member
    edited February 2013

    I'd like to restart mine every hour or so, keeps me busy.
    Not to mention complete re-installs every week, just as an excuse not to go out.

  • @ErrantWeb said: Time for a reload. :P


    @shovenose said: Or a new laptop.

    Any $7 laptop around? :P

    On topic, I rarely rebooting servers. Just restarting a few services like mysql occasionally. That way I can still show off uptime output and grow my penis.

  • great, thanks for all your info guys, you are awesome!

Sign In or Register to comment.