Record uptimes

Johnny2Bad

Johnny2Bad

I have been mucking around with FreeBSD 8.2 for quite some time and I was just curious as to how long the average user manages to keep FreeBSD up for. My record is a little over 24 days. I managed to get the thing up for 19 days this time, but in both cases GNOME has stalled, totally locked up the system which has caused me to power down.

To address the above issue, I was thinking of only running GNOME from the command prompt and only use it when I'm feeling lazy or in need of some graphical interface. But the question remains, what is the longest uptime anyone has experienced with FreeBSD?

Thanking you in advance,
Jonathan.
 
Uptime is a worthless statistic. A single-minded FreeBSD server (without X) can stay up for years on end, but it's a security hazard after a couple of weeks/months (when on a public network). A FreeBSD workstation (with X) should be able to stay up a very long time as well (the same security hazard warning applies), though there is a little more risk of starving resources due to the much larger amount of running applications, each with their little memory leaks, etc. I have never seen a lock-up on my FreeBSD laptop, though I don't keep it running more than a day, and I don't run behemoths like Gnome/KDE. I can see how those might cause problems after a while. Anyhow: don't aim for indefinite uptimes, aim for a secure system.
 
Thank you for pointing this out.

However with my current setup I think I will stick to only running GNOME the behemoth as you say because I use my FreeBSD server as a workstation too. But as I said I will go back to running GNOME from the command prompt and exiting it when I've finished with it in its workstation role.
 
I've had a server run almost 300 days which I neglected to patch years ago. At one point netcraft had FreeBSD in the longest uptime statistic which was years. Maybe still does.

Either way if you have downtime its usually a configuration error. I know zfs was causing me some problems with server reboots at it's initial release( though I also mated it with WD green drives which made it a double whammy).
 
There's a famous quote going around of Linus Torvalds saying: "How do you power off this machine?". Whether it was exaggerated/folklore or not, that machine apparently hadn't been updated in ages. Whether or not that's a good thing (if only security-wise) is left as an exercise to the reader (I've always wanted to say that!).

Fonz
 
FreeBSD in it's "flab free" mode with no X window system and no Desktop Environment, running robust services like DNS, DHCP, FTP etc; should be able to run pretty much until the power is cut. Similar with my Ubuntu home server my FreeBSD home server can easily go past 100 days and not break a sweat.

Up time though is a very poor metric of Server performance/reliability; as already mentioned, a server that has not been restarted in that time has not been patched to the latest kernel errata and could be easily on the verge of a failure. For a desktop machine this isn't really important to worry about, but for a production server you should adhere to maintenance windows, patch, update and restart the box when it's due.

There is nothing more scary than the box in the corner that's got 400-1000 days of up and no one knows how to fix it if it crashes.

Alas, it's very cool to see some of these big numbers. Sadly though it seems most of the highest up time boxes are now Windows Servers; Hey, maybe Unix/Linux admins are now getting savvy to good practice :pP - God have mercy on us.

P.S OS X on my laptop will run anywhere from 30-45 days before it grinds to a halt.
 
I had a FreeBSD 4.x box run for several years. It was simply doing IPSEC (for a point to point WIFI link) and SQUID on a remote site - not visible on the internet, only reachable via our corporate WAN).

I screwed up an SSH upgrade way back, and never ended up getting to the remote site before I left the company.

I came back 2 years later and it was still running just fine (with zero maintenance) - the only outages it had experienced were due to power cuts.

Typically I get 180 days or so out of my boxes that are well protected on our internal network running DNS/sendmail. The only time they go down is due to kernel upgrades or scheduled downtime for server relocation (our rack space is tight).

Generally, I think you'll find the need for reboot due to kernel security patch or some other external requirement (power, etc) will come about before FreeBSD needs to be rebooted.


One thing I will add: if you are making major configuration changes to a box, it is a good idea to schedule a reboot afterwards to ensure that your box still boots properly and all services start. There is nothing worse than having a box go down due to some other issue (power, scheduled restart, etc) and then not start properly due to changes you did 6-12 months ago.

When I was working in an ISP, we had an admin who was notorious for making changes and not altering the startup RC scrips to reflect the changes made. Hence, the box would run for months at a time until a reboot was performed at which point random stuff would break.
 
This is uptime for one old server... server running with bad sectors and reboot would probabliy kill it...
Screenshot from 2016-01-06 10:16:38.png

Screenshot from 2016-01-06 10:17:31.png
 
I hope that thing isn't connected to the internet. If it is it's probably not your server any more.
 
Well, spewing spam and malware for about 10 years would be one hell of a testsuite, would it not? Quick, someone tell Phoronix ;) But SirDice is right, it will be owned ten ways sideways to sunday if it is reachable from the outside.
 
Code: 
# uptime
10:16AM  up 114 days, 22:04, 18 users, load averages: 0.33, 0.31, 0.33

10.2-RELEASE FreeBSD 10.2-RELEASE #0
desktop with kde
 
SirDice said:
Patch that thing! As in right now.
Click to expand...

There are the cross-overs on youtube with a scene from "Der Untergang" where new subtitles are added to a scene in which Adolf goes nuts. Is there anyone able to do a mashup from "Full Metal Jacket"? Sgt Hartman yelling in your ear "Patch that thing!" - at least that is the mental image with which my mind came up with when I read your post :D

Edit: That might start a new meme, so glory and fame await the noble artist who does the first one ;)
 
I have observed some FreeBSD servers running for a few years at a time, I think around 2.5 or 3 years was the max I have seen in person. My own server machines are usually rebooted about once per quarter (ie 90 days) for security patches and to make sure they boot cleanly.
 
NewGuy said:
and to make sure they boot cleanly.
Click to expand...
That's probably the biggest reason why I regularly reboot machines. Admins shouldn't be afraid to reboot servers. Especially after you made changes to its configuration or boot parameters. More than once did I come across a server somebody worked on but never checked if it actually boots correctly. Great when you have to clean up after a power outage or you reboot after installing patches. Then suddenly the machine refuses to boot.

Moral of the story, if the server cannot be rebooted because of some application it's running you're doing something wrong.
 
Even as a full bloat KDE desktop machine, I haven't had any lockups. I bet it would stay up forever if I didn't reboot every time I do an update.
 
You must log in or register to reply here.
Back
Top