What are some of the known longest-uptime *nix installs in the world?
June 30, 2015 1:06 PM Subscribe
Someone mentioned that netcraft used to have an updated ranking, but no longer does.
I imagine this would mean no kernel updates or anything for many years too, so I guess it would be a pretty un-secure machine.
The longest *nix uptime I can find is a Sun 280R server running Solaris 9 that ran for 3737 days = 10ΒΌ years.
Even longer, though, is this NetWare 3.12 server that lasted 16 years.
posted by Rangi at 1:16 PM on June 30, 2015
Even longer, though, is this NetWare 3.12 server that lasted 16 years.
posted by Rangi at 1:16 PM on June 30, 2015
Response by poster: I originally had 'current' in the title but removed it, stupidly, since I actually meant CURRENTLY running - as of today.
posted by basehead at 1:21 PM on June 30, 2015
posted by basehead at 1:21 PM on June 30, 2015
I think the longest I've ever personally managed on bare metal is somewhere around 2-3 years on some crazily outdated Ubuntu release (a box that I would be unsurprised to find lives on, in that configuration, as a VM still printing many thousands of labels a day).
It doesn't seem crazy to think that there's (relatively) a lot of stuff out there in the 3-5 year range, even using desktop-oriented Linux distributions and middling consumer-grade hardware. I'm gonna guess it falls off dramatically after that, in the zone where catastrophic long-term power failure and the half-lives of employable sysadmins, viable businesses, and grad student careers really start to cast a shadow.
posted by brennen at 1:24 PM on June 30, 2015
It doesn't seem crazy to think that there's (relatively) a lot of stuff out there in the 3-5 year range, even using desktop-oriented Linux distributions and middling consumer-grade hardware. I'm gonna guess it falls off dramatically after that, in the zone where catastrophic long-term power failure and the half-lives of employable sysadmins, viable businesses, and grad student careers really start to cast a shadow.
posted by brennen at 1:24 PM on June 30, 2015
Since nobody else has provided definitive proof of a long uptime, I'll set the bar with my workhorse SQL server, which is actively providing the database backend for several dozen Wordpress, Joomla, Wiki, and other websites accessible from the internet:
azraelbrown@sql:~$ uptime
15:36:58 up 1326 days, 3:02, 1 user, load average: 0.11, 0.06, 0.05
Yes, I should patch it, but it's not worth the downtime and the server just runs and runs and runs and never gives me any trouble...
posted by AzraelBrown at 1:40 PM on June 30, 2015 [1 favorite]
azraelbrown@sql:~$ uptime
15:36:58 up 1326 days, 3:02, 1 user, load average: 0.11, 0.06, 0.05
Yes, I should patch it, but it's not worth the downtime and the server just runs and runs and runs and never gives me any trouble...
posted by AzraelBrown at 1:40 PM on June 30, 2015 [1 favorite]
I once had an older model bit of hardware from F5 that ran BSD on commodity Dell hardware and it had north of 8 years of uptime last I heard, which I thought spoke as well to the AC power availability in the room as it did the hardware/OS itself. And yeah, it was running under load to boot.
posted by jquinby at 2:38 PM on June 30, 2015
posted by jquinby at 2:38 PM on June 30, 2015
4540 days for a bsd machine that is still in use. Techs know to tread lightly around that box.
posted by Radiophonic Oddity at 2:40 PM on June 30, 2015 [1 favorite]
posted by Radiophonic Oddity at 2:40 PM on June 30, 2015 [1 favorite]
Ridiculous uptimes will become more common soon since a ksplice workalike is supposed to merged into mainline Linux soon, if it hasn't already. This will allow for kernel patches without a reboot on systems running distribution default kernels. Given that is pretty much the only reason to actually reboot a Linux server, expect most of them to run until they have a hardware or power failure.
Not current, but I had a client that had an AIX system that had an uptime well over 6 years when we decommissioned the server. (It ran a Muzak distributor's billing software)
Right now I don't have any that have been running for more than 150 days since there was a remotely exploitable kernel bug I had to reboot all my servers for a few months back. :(
posted by wierdo at 3:17 PM on June 30, 2015 [2 favorites]
Not current, but I had a client that had an AIX system that had an uptime well over 6 years when we decommissioned the server. (It ran a Muzak distributor's billing software)
Right now I don't have any that have been running for more than 150 days since there was a remotely exploitable kernel bug I had to reboot all my servers for a few months back. :(
posted by wierdo at 3:17 PM on June 30, 2015 [2 favorites]
There will be DEC VMS installations doing industrial things with uptimes from the 90s. They just will never be public facing or reported.
posted by scruss at 3:56 PM on June 30, 2015 [4 favorites]
posted by scruss at 3:56 PM on June 30, 2015 [4 favorites]
Apparently there was a 20-year Tandem Nonstop installation at a bank. I think the product like got killed before the actual computer had to be shut down ironically.
posted by GuyZero at 3:59 PM on June 30, 2015
posted by GuyZero at 3:59 PM on June 30, 2015
Not exactly what you're looking for, but I had a Solaris 2.7 server that I booted shortly after starting at a former workplace and shutdown shortly before I left, 5 years later.
posted by jferg at 6:05 PM on June 30, 2015
posted by jferg at 6:05 PM on June 30, 2015
Ridiculous uptimes will become more common soon since a ksplice workalike is supposed to merged into mainline Linux soon, if it hasn't already. This will allow for kernel patches without a reboot on systems running distribution default kernels.
Also, (a) clusters and (b) VM migration. I guess it depends on how you define an 'install' or 'host.'
Thinking about it, might it be likely that a satellite with some kind of embedded *nix holds the current lead? Or are all the RTOSes used in space POSIX but not *nix?
posted by snuffleupagus at 8:26 PM on June 30, 2015
Also, (a) clusters and (b) VM migration. I guess it depends on how you define an 'install' or 'host.'
Thinking about it, might it be likely that a satellite with some kind of embedded *nix holds the current lead? Or are all the RTOSes used in space POSIX but not *nix?
posted by snuffleupagus at 8:26 PM on June 30, 2015
To answer my own question, it seems that putting *nix on a spacecraft is something that is just now being attempted (the Langely cubesat.) So that's not it.
posted by snuffleupagus at 8:37 PM on June 30, 2015
posted by snuffleupagus at 8:37 PM on June 30, 2015
There are a lot of anecdotes about AS/400 systems pushing 3,000+ days of uptime. They are almost always in some sort of weird isolated role, e.g. industrial process control or something, not internet-facing, sometimes not even connected to TCP/IP. Very difficult to verify.
Here is a thread about it, with some stories, from the MIDRANGE-L list. (And the usual wankery you get in any uptime discussion, i.e. potshots at "Windoze", people jumping in to lecture everyone on why uptime doesn't matter, etc. etc.)
One person mentions a PDP-11 that supposedly ran for 14 years without a reboot. No substantiation on that, though, and there might be other systems from that era with similarly long uptimes, given how involved a process rebooting them was.
posted by Kadin2048 at 9:33 PM on June 30, 2015 [1 favorite]
Here is a thread about it, with some stories, from the MIDRANGE-L list. (And the usual wankery you get in any uptime discussion, i.e. potshots at "Windoze", people jumping in to lecture everyone on why uptime doesn't matter, etc. etc.)
One person mentions a PDP-11 that supposedly ran for 14 years without a reboot. No substantiation on that, though, and there might be other systems from that era with similarly long uptimes, given how involved a process rebooting them was.
posted by Kadin2048 at 9:33 PM on June 30, 2015 [1 favorite]
I dunno... I don't think you'd find that many *current* examples listed anywhere; the world is increasingly scary for elderly, unpatched machines, so IMHO few sysadmins/orgs would go crowing about it in public.
(Source: I'm a sysadmin working for a large org and, even though you don't know *which* org I work for (at least I hope you don't...!), I'm still not going to tell you how long our longest-running unix machines have been up... :) )
posted by snap, crackle and pop at 10:42 PM on June 30, 2015 [1 favorite]
(Source: I'm a sysadmin working for a large org and, even though you don't know *which* org I work for (at least I hope you don't...!), I'm still not going to tell you how long our longest-running unix machines have been up... :) )
posted by snap, crackle and pop at 10:42 PM on June 30, 2015 [1 favorite]
I also came in to mention KSplice (now owned by Oracle) allowing Linux to reload the kernel without a reboot -- and thereby making Uptime Heroics a thing of the past. :7(
posted by wenestvedt at 9:29 AM on July 1, 2015
posted by wenestvedt at 9:29 AM on July 1, 2015
This thread is closed to new comments.
posted by brennen at 1:15 PM on June 30, 2015