We have about 6 linux boxes running for about 8 months straight now. Should we restart them? There are no problems, but how often do you guys restart your Linux web servers?

I just leave mine on till they have to be restarted, the last box I restarted was 28 days ago and that was because of a memory upgrade....

if it ain't broke don't fix it


I have heard of linux boxes going for over a year with out a reboot. Generally I only reboot the server if there is a problem or some work was done on it that needs a reboot like a kernal upgrade.

dherman76,

It depends on what you take into consideration.

If you have demanding sites that use alot of php and mysql, you may need to restart your machines to keep your resources free.

I have redhat and freebsd servers with 512mb ram in each, even though there may not be any apparent problem with the server, I tend to restart my servers evey month to keep the resources free and high.

Tcoy, and The ProHacker are also correct, most of the times servers are only reboot when there are problems with them or there are upgrades going on.

Well as already stated leave it up till it needs to be rebooted...

Some just reboot once a month to have a clean slate but far as I care no need to mess with it. My server at catalog was push 60+ days before their noc lost power. probabbly would still be up running with no problems... :D

It really depends on what's on the server. For most people, letting it go until it's necessary (power outage, hardware upgrade, software install upgrade that requires reboot, and so on) is fine.

For instance, across eight servers, we have uptimes of:
11:37am up 68 days, 22:40
11:41am up 9 days, 8:41
11:42am up 37 days, 9:17
11:42am up 18 days, 12:41
11:43am up 8:49 (rebooted today for secure cert)
8:43am up 13 days, 22:00
8:44am up 4 days, 12:54 (up since it was brought online)
The system has been running for 0 days, 23 hours, and 33 minutes (this is a RaQ that we've picked up just today from someone we bought out)

My personal preference is to just leave things alone unless there is some compelling reason to mess with them.

Originally posted by Annette


My personal preference is to just leave things alone unless there is some compelling reason to mess with them.

Annette is 100% correct. I have had a linux web server that ran for 240 days before I had to reboot it.

I wonder whats the longest for a win/nt server... 36 hours maybe? :D

Originally posted by Madman2020


Annette is 100% correct. I have had a linux web server that ran for 240 days before I had to reboot it.

I wonder whats the longest for a win/nt server... 36 hours maybe? :D

Now, now, let's not be too terribly harsh on the NT/2K side of things. Properly managed, they can have some pretty decent uptimes as well. I babysit a whole bunch of them, and with one exception, they all have uptimes of +15 days, with about half going +30 and a quarter going +50. There is one that that has to be rebooted three times a day, but I'm convinced that it's Domino (yeccchhh) and not the NT install.

We rarely reboot our UNIX (BSD mainly) servers. Usually, the only things that prevent 300+ day uptimes are hardware upgrades. We have 5 or 6 that have 100+ day uptimes currently.

We generally end up having to reboot our NT servers every ~60 days. We don't normally have any problems before then, though.

No Win2k servers, yet.

Originally posted by Annette


Now, now, let's not be too terribly harsh on the NT/2K side of things. Properly managed, they can have some pretty decent uptimes as well. I babysit a whole bunch of them, and with one exception, they all have uptimes of +15 days, with about half going +30 and a quarter going +50. There is one that that has to be rebooted three times a day, but I'm convinced that it's Domino (yeccchhh) and not the NT install.

My opinion: NT/2000 sucks when it comes to stability. 2000 is excellent, full of features, probably the very best platform for a large LAN environment, but as far as stability is concerned ----> :bawling:

Unix based and others such as Novell Netware are very stable. We have a few Novell boxes on our LAN running for about 200 + days. Our 2000 boxes run 20 to 30 days? :D

I used to have a link to a website that maintained records of the longest uptime's for Linux/Unix boxes, but I'll be dammed if I can find it. When I last looked, the longest uptime was 1300+ days.

We have some Windows 2000 boxes that have gone for 6 - 8 months without a reboot.

It all depends on how the box is managed and configured.

Our NT DNS servers have been up for 8 months now since the last reboot (knock on wood).

Unix-like systems generally only have to be rebooted when the hardware is upgraded. I know this guy who has been running a box for 3+ years without a reboot, now he won't even upgrade it because he's aiming for 5 years. Unless he has a hardware failure, I'd say he'll make it. I think he's mad. :eek:

Originally posted by avara
Unix-like systems generally only have to be rebooted when the hardware is upgraded. I know this guy who has been running a box for 3+ years without a reboot, now he won't even upgrade it because he's aiming for 5 years. Unless he has a hardware failure, I'd say he'll make it. I think he's mad. :eek:

He has a UPS hooked up to I take it? The batteries on the UPS have to wear out somewhere in that 5 year time frame. ;)

Originally posted by TradeViceroy


He has a UPS hooked up to I take it? The batteries on the UPS have to wear out somewhere in that 5 year time frame. ;)

Good point!

I'll be sure to point this out to him. :cool: Seriously though, I think he's insane... Nobody with a sane mind would get a kick out of running a server for 5 years without a reboot, right? :D

Originally posted by avara


Good point!

I'll be sure to point this out to him. :cool: Seriously though, I think he's insane... Nobody with a sane mind would get a kick out of running a server for 5 years without a reboot, right? :D

Are you mocking my ultimate purpose in life? :mad:

Well it really depends,
see, if it aint broke dont fix it IS a good remark...
But we have an auto rebooter for the "heavy" machines, they auto reboot once a day... Just to give them a clean start :)

In a "Network" situation a daily reboot is ok, and often does wonders to clear stored cached memory, etc. But in the "Internet" world a reboot means downtime.

Kind of fell the same way.. If it aint broke dont fix it IS a good remark... :)

I have some that have been up for about 6 months some a month but no need really if there is no problem..

We have a local ISP here and they reboot their WindowsNT servers every 12-24 hours :nuts: :D

Well, my PC is not a server, and I reset it at least 4 or 5 times daily (WinMe) :mad:

WinME really sucks.. Win 98 was much better..

Originally posted by jokerxixlx
WinME really sucks.. Win 98 was much better..

Agreed! :angry:

I would never even consider using WinMe nor 98. They both suck big time. The Microsoft people just keep trying to keep these operating systems backwards compatible with 16 bit applications, which makes them evil and probably the very worst operating systems.

I use 2000 Professional and it is much more stable, still not the best, but much better. I only have to reboot my Win 2000 PC every 2 or 3 weeks. :D

When i used to have Windows98, it crashed every 2-3 hours (maybe i'm working too much :D). I also have Windows2000 Professional, which unfortunately doesn't supports one of the newest CD-DVD Drives :rolleyes: . Now i have WindowsME - running since 3 days without a reboot (800Ghz Processor, 512MB RAM) and it works just fine :) Oooh, and i'm also running SuSe Linux on the other partition :D

Originally posted by jokerxixlx
WinME really sucks.. Win 98 was much better..

Win95c was the best. Win98SE was a little better that 98, which was pants.

Me is ok, but it's a bit of a cobble between 98SE and 2k, and doesn't go too well. 2K is pretty good, but needs much more memory than NT4.

I still use 95C. As I don't need USB, it works perfectly :).

Rich

So far i have been happy with win2k.. Much better then and 98 or me will ever be.. :)

Originally posted by jokerxixlx
So far i have been happy with win2k.. Much better then and 98 or me will ever be.. :)


Yup. w2k is easily the best OS that Microsoft has ever sold (though, that is not saying a lot). I use it as my main PC at home.

It still has the typical MS quirks. My registry settings for dial-up networking have a habit of changing themselves, ditto for the Windows Explorer folder options.

And I still have to reboot it each evening so that it doesn't lock up every couple of weeks. It is better than NT 4.0SP6a in this regard, that OS used to lock up every 2 or 3 days on me.

On the other hand, my OpenBSD box just runs and runs and runs. :)

OpenBSD and FreeBSD are great you can't beat them for nothing..

I tested out WIN XP man i have to say it really sucks Win2k was much better really..

I think i will stay with win2k for some time works and only have to reboot about once a month or so so i can live with it..

But just my home system has a gig of ram and dual 800 PIII's great for games and stuff.. :)

<<Admin edit : cross-posted>>

used to have a link to a website that maintained records of the longest uptime's for Linux/Unix boxes, but I'll be dammed if I can find it. When I last looked, the longest uptime was 1300+ days.

That might just be this link:
http://uptime.netcraft.com/up/today/top.avg.html

Pretty scary being up for 3+ years...

Jan Derk

And what is nice they are all FreeBSD... LOL Not one MS server even there... LOL Hell not even a redhat.. hehe

10:40pm up 442 days, 20:23, 0 users, load average: 0.00, 0.00, 0.00

It's all about Debian. :D

my max is 6 months uptime for an NT server - and we had a power failure and the UPS shut it down :bawling:

not many people will have a very high uptime for an IIS server, afterall dont you need to reboot it to implement the patches for all the bugs that come out?? :(

Yea most of the time for patchs you have to reboot it kind of makes you.. 6Months is great for any microsoft software..lol

I have been on servers with over a 1000 day uptime that are running linux (over 3 years).. There is nothing wrong with them and they run fine.

Originally posted by Vince
10:40pm up 442 days, 20:23, 0 users, load average: 0.00, 0.00, 0.00

It's all about Debian. :D

Okay, what did you do with that box? Just leave it in the basement running for 1.5 year, so you can say it is up 442 days right now? :D

Really, our server is having major troubles for some reason I can't explain... it has a stable server load (0.10 - 0.20) usually, but once every day, it goes to over 200.00 (yes, that's high) , I can't telnet to it, and our server is about half the world away from me (I'm in Europe, server in the USA) , so I can't reboot it. So I have to ask my webhost every day to reboot it (which usually takes them 15 minutes) ...

The longest uptime we had the last month was around 3 days... on a RH7.0 box

Originally posted by blazenet
but once every day, it goes to over 200.00

Somethings majorly screwy with your box, it seems. I'd have someone who knows RH (hopefully your host) investigate what may be causing this.

100 day uptimes shouldn't be a big thing for unix-based servers.

5:12PM up 10 days, 20:38, 1 user, load averages: 0.01, 0.01, 0.00
That is my FreeBSD box here at home, I just set it up about 10 days ago ;)


8:06am up 47 days, 20:27, 1 user, load average: 0.18, 0.05, 0.01
And that is my RaQ3 box at Tera-Byte, which I've had for 7 months now.

Originally posted by Planet Z


Somethings majorly screwy with your box, it seems. I'd have someone who knows RH (hopefully your host) investigate what may be causing this.

100 day uptimes shouldn't be a big thing for unix-based servers.

The problem is, our web host charges $150 / hour... We don't really have that much money to spend, so we're in kind of trouble...

Getting a new server would be a better (and cheaper) solution...

Is it a dedicated server? (not a colo). Then they really shouldn't be charging for it.

Your best bet may be, as you said, just to find a new server/host. Let your old host know that if they can't/won't fix it for free, you'll find a better place.

We reboot our machines if we are installing some software patches, upgrading the kernel, etc. other than that we don't reboot the servers.

Originally posted by Planet Z
Is it a dedicated server? (not a colo). Then they really shouldn't be charging for it.

Your best bet may be, as you said, just to find a new server/host. Let your old host know that if they can't/won't fix it for free, you'll find a better place.

Yes it is a dedicated server, so my feeling is, we rent something from them, it malfunctions, and we have to PAY to fix it?

That sounds unlogical to me, but I guess if they say it is our fault, we can't do nothing.

Originally posted by blazenet
That sounds unlogical to me, but I guess if they say it is our fault, we can't do nothing.

Except move. ;)

Yeah...

Any recommendations for a good FreeBSD (I was thinking of switching to FreeBSD a long time, because I heart a LOT of good stories about it) webhost, which is reliable and has good (fast) support? Preferably ICQ support and such...

UUNet is too expensive for me, and I believe RackSpace guarantees a max of 7 seconds downtime / month... that caught my interrest ;)

Originally posted by blazenet
Really, our server is having major troubles for some reason I can't explain... it has a stable server load (0.10 - 0.20) usually, but once every day, it goes to over 200.00 (yes, that's high) , I can't telnet to it, and our server is about half the world away from me (I'm in Europe, server in the USA) , so I can't reboot it. So I have to ask my webhost every day to reboot it (which usually takes them 15 minutes) ...

I don't understand? Is it not possible to telnet to a server no matter where it is located on the internet if the ISP/Host makes the telnet option available?

So why can't this particular server be rebooted from another continent?

Another thing I'll be learning here today!

the computer is too "overloaded" to allow telnet sessions when at a load average of say 200..

if u wanna put it that way/maybe that'll help you understand.

Greetings:

We have some Linux and Sun boxes that have not been rebooted in over two years.

We reboot for hardware upgrades.

Other than that, it is on a case by case basis with the idea of trying not to reboot unless absolutely necessary (which on Unix is easy).

Thank you.

I have over 20 months of uptime on the windows 2000 server that I manage for the company I work for... I'll have to check on exact time this evening. This server is used for our Intranet only, and we use it for the Active Directory support, which is better than Novell by 100x.

I checked our server here at school... We have eight months off of it... Its a Windows 2000 Server box running Novell 4.

My home computer, running 2000 Advanced Server as well, gets close to three months before I have to reboot. Mind you, though, I use it for gaming, software development, etc. and it only has 256 megs of RAM, so it slows down after a while. :cool:

Windows is not that bad... a reboot every six months isn't bad.

I have never taken IIS offline, even to implement patches. I just install it to another copy, map another IP to it, and just switch the router's forwarding profile to the other IP... At most, three seconds of downtime.

I use it for the ASP and ASP.NET support and for SQL Server. mySQL is great, but SQL Server, for enterprise applications, it smokes mySQL in almost any tests. I just prefer its environment.

*nix is better... I'll agree, but Windows has some features you just can't get on nix.

- James

Originally posted by dynamicnet
Greetings:

We have some Linux and Sun boxes that have not been rebooted in over two years.

We reboot for hardware upgrades.

Other than that, it is on a case by case basis with the idea of trying not to reboot unless absolutely necessary (which on Unix is easy).

Thank you.

That sort of attitude is nonsense. This means that you haven't upgraded your kernel to the new and stable, more bug free versions for over 2 years. It's not voodoo to reboot your server, you know. Something more reasonable, would be; "We have never needed to reboot a server and haven't had a server crash for over 2 years".

Greetings Tim:

"That sort of attitude is nonsense. This means that you haven't upgraded your kernel to the new and stable, more bug free versions for over 2 years."

1. We have Sun Solaris boxes and Linux. Sun doesn't update the Kernel that often.

2. If it is not broke, then do not fix it ;-)

Any questions? See #2.

Thank you.

Originally posted by dynamicnet
Greetings Tim:

"That sort of attitude is nonsense. This means that you haven't upgraded your kernel to the new and stable, more bug free versions for over 2 years."

1. We have Sun Solaris boxes and Linux. Sun doesn't update the Kernel that often.

2. If it is not broke, then do not fix it ;-)

Any questions? See #2.

Thank you.

Yeah, who needs a stable, secure kernel version anyway, right? Great advice.

Greetings Tim:

"Yeah, who needs a stable, secure kernel version anyway, right?"

Sometimes the latest is not the most stable.

Whose to state the release we have now is not stable?

What happened to the rule of thumb, "if it is not broken" don't fix it?

In any event, we can agree to disagree.

Take care.

Originally posted by dynamicnet
Greetings Tim:

"Yeah, who needs a stable, secure kernel version anyway, right?"

Sometimes the latest is not the most stable.


I promise that a kernel that's two years old, is not secure or stable.


Whose to state the release we have now is not stable?


You said you haven't rebooted in 2 years, I am "who" will state that your kernel is therefore not secure or stable.


What happened to the rule of thumb, "if it is not broken" don't fix it?

In any event, we can agree to disagree.

Take care.

Yeah, and all you do it disagree, no matter how logical or factual the argument against your nonsense posts are. I'd have to say you are possibly a liar to boot. Nonetheless, you are on ignore, because I can't stand a hypocrite (or a liar). However, stop giving bad and dangerous advice out, like it's "fine" to do, because it's not.

linux is not meant to be rebooted often, if you reboot it say...every other night, it will lead to serious problems with alot of stuff

Greetings Tim_Greer:

Are you stating the Kernel's released by NTT/Verio, CommuniTech (now Interland), EarthPort, Rackshack.net, and many others are insecure because they are not the latest?

Also, as I stated in my earlier post, we have several Sun's. They don't upgrade Sun's that often.

But in any event, you are choosing what to read and not read; what to assume and what to not assume.

Take care.

Originally posted by Softicom.NET
linux is not meant to be rebooted often, if you reboot it say...every other night, it will lead to serious problems with alot of stuff

Just to be on the safe side, about my previous post; I also don't recommend rebooting all the time, or even rebooting like a lot of people do, to "fix problems". I.e., something is using too much server load, there's an error, etc., I see a lot of people reboot their server to fix it. However, rebooting isn't the end of the world, and I only do it (personally) when there's something, like a kernel upgarde, that requires it. So, you rarely should need to reboot it.

Greetings Tim:

"So, you rarely should need to reboot it."

We are in agreement.

You are correct.

Thank you.

Originally posted by Softicom.NET
linux is not meant to be rebooted often, if you reboot it say...every other night, it will lead to serious problems with alot of stuff

such as?

could you give one or two examples please?

Originally posted by Chang Lee


such as?

could you give one or two examples please?

Unless you have a corrupted disk, or some other faulty component, rebooting a Linux machine won't cause any problems at all. Heck just for fun, I just rebooted by Linux machine at home 10 times in a row, and it still appears to be working fine. :cool:

But back on topic, I do not think it is necessary to reboot Linux often. Apart from kernel upgrades or running fsck, most patches and upgrades do not require the entire machine to be rebooted. So depending on the configuration, 100+ days of uptime usually isn't a problem at all, and 300+ shouldn't really be a major problem either.