Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[TheBoy]

Blimey, logged on to one at work today, and I always run a few standard commands to check various bits, irrespective of the fault/planned work which is why I'm on.

One today had an uptime of 6.8yrs

This is a truely internet facing server that a fair few of UK internet users will use.  I guess its dropped off the patching lists


But 6.8yrs for a fairly busy server is still amazing

[cem_devecioglu]

6.8 years ..    I bet its not a windows based server

[TheBoy]

6.8 years ..    I bet its not a windows based server

Err, no.

That said, my brother had an NT4 server, on a private network, manage 5yrs uptime

[Martian]

6.8 years ..    I bet its not a windows based server

Err, no.

That said, my brother had an NT4 server, on a private network, manage 5yrs uptime

Many moons ago I had a 2K server (also running Exchange and a couple of other bits of mundane software) stuffed in the rack at Redbus for just under 4 years running 24/7 no problem. I still have the pics somewhere of the day I took it out of the rack after all that time running, and it was as spotlessly clean inside as the day I built it.

As far as Windows boxes being unreliable, they really only go wrong when the owner starts installing poorly written 3rd party crap.

[TheBoy]

6.8 years ..    I bet its not a windows based server

Err, no.

That said, my brother had an NT4 server, on a private network, manage 5yrs uptime

Many moons ago I had a 2K server (also running Exchange and a couple of other bits of mundane software) stuffed in the rack at Redbus for just under 4 years running 24/7 no problem. I still have the pics somewhere of the day I took it out of the rack after all that time running, and it was as spotlessly clean inside as the day I built it.

As far as Windows boxes being unreliable, they really only go wrong when the owner starts installing poorly written 3rd party crap.

Couldn't agree more, although MS have been known to push out the occasional product that crashes things - less so in the server arena though.

Windows Server has 2 problems:
It's the same (but tweaked) kernel as desktop OS, which means it gets targetted by baddies.
It's the same (but tweaked) kernel as desktop OS, so any Tom, Dick and Harry can write substandard software that will run on the server OS.


Its very, very rare to see a Windows Server fall over nowadays.  TBH, not really since the days of NT 3.51 and NT4, although these were always as a result of duff drivers.  In fact its rare to see any Windows machine fall over, if it does its usually a result of crappy unsigned, backstreet drivers, or failing hardware.


My day job is TSS for thousands of servers, covering mostly Windows, Linux, Solaris and a little HP-UX.  Sticking with Windows for a moment, we tend to run this either directly on HP Proliant DL server hardware, or via a hypervisor running on HP Proliant BL blades.  This combination tends to give us very, very few problems, as the hardware is pretty robust, the drivers are tried and tested, and Windows is rock stable.

We tend to run Linux on the same hardware, not as stable, partly due to less mature drivers available for Linux, made worse by a licencing problem if drivers taint the kernel (many do).


The machine in the original post was an elderly Sun Sunfire 280R, running Solaris 8.  Shame Oracle (who bought Sun) can't make the current hardware reliable - I've spent all this week trying to build 5 Oracle blades across 3 chassis. 2 chassis had firmware issues that were not fixable by end user (ie, me) as Oracle don't make the firmwares available, 4 faulty blades, and now a faulty network module. FFS.

[VXL V6]

Biggest problem I ever see with unix servers is when some halfwit support person decides to solve an 'issue' by issuing a kill -9 command on the process, thereby leaving a zombie, usually a few hundred times over a period of time.

 

[TheBoy]

Biggest problem I ever see with unix servers is when some halfwit support person decides to solve an 'issue' by issuing a kill -9 command on the process, thereby leaving a zombie, usually a few hundred times over a period of time.

 

Aye, if you're gonna do a kill -9, at least do a ptree first to kill the hung process, not a parent.

[VXL V6]

The machine in the original post was an elderly Sun Sunfire 280R, running Solaris 8.  Shame Oracle (who bought Sun) can't make the current hardware reliable - I've spent all this week trying to build 5 Oracle blades across 3 chassis. 2 chassis had firmware issues that were not fixable by end user (ie, me) as Oracle don't make the firmwares available, 4 faulty blades, and now a faulty network module. FFS.

Ah yes, the original Sun Hardware I used to look after (Sparc stations and servers) was rock solid in it's construction, I bet the Type 4 and 5 keyboards are still going now compared to the utter s_____ that gets shipped with any brand of PC now.

Mind you, I bet you wouldn't like the Sony Trinitron VDU's they used  , I always seemed to end up having to lug them up eight flights of stairs to some design office then end up having to swap their other screens out as well because the screens were different shades.

[albitz]

Absolutely.



 

[VXL V6]

Biggest problem I ever see with unix servers is when some halfwit support person decides to solve an 'issue' by issuing a kill -9 command on the process, thereby leaving a zombie, usually a few hundred times over a period of time.

 

Aye, if you're gonna do a kill -9, at least do a ptree first to kill the hung process, not a parent.

I should have said 'saw' as I don't really have much to do with Sun kit now, only a few Sun (Well Oracle now) PC type servers running Terminal Services and Sunray back ends and a few hundred Sunray thin clients.

[TheBoy]

The machine in the original post was an elderly Sun Sunfire 280R, running Solaris 8.  Shame Oracle (who bought Sun) can't make the current hardware reliable - I've spent all this week trying to build 5 Oracle blades across 3 chassis. 2 chassis had firmware issues that were not fixable by end user (ie, me) as Oracle don't make the firmwares available, 4 faulty blades, and now a faulty network module. FFS.

Ah yes, the original Sun Hardware I used to look after (Sparc stations and servers) was rock solid in it's construction, I bet the Type 4 and 5 keyboards are still going now compared to the utter s_____ that gets shipped with any brand of PC now.

Mind you, I bet you wouldn't like the Sony Trinitron VDU's they used  , I always seemed to end up having to lug them up eight flights of stairs to some design office then end up having to swap their other screens out as well because the screens were different shades.

We have very few Sun keyboards and VDUs here, as they are extras.  99.9% of our stuff is rack mounted with a serial connection up its jacksie (or LAN to NetMgmt port if its from this millenium )

[Ian_D]

6.8 years 

Thats bloody impressive! 

Dunno how long my home server has been running since last reboot, will check and see in a moment. However it has been turned on 24/7 (minus the few hrs when its been off for hardware changes / UPS upgrades etc) for almost 5 years. Its only a crappy IBM eServer too! x206 IIRC... Which reminds me, need to shove a new backup tape in - its VERY VERY overdue 

[VXL V6]

The machine in the original post was an elderly Sun Sunfire 280R, running Solaris 8.  Shame Oracle (who bought Sun) can't make the current hardware reliable - I've spent all this week trying to build 5 Oracle blades across 3 chassis. 2 chassis had firmware issues that were not fixable by end user (ie, me) as Oracle don't make the firmwares available, 4 faulty blades, and now a faulty network module. FFS.

Ah yes, the original Sun Hardware I used to look after (Sparc stations and servers) was rock solid in it's construction, I bet the Type 4 and 5 keyboards are still going now compared to the utter s_____ that gets shipped with any brand of PC now.

Mind you, I bet you wouldn't like the Sony Trinitron VDU's they used  , I always seemed to end up having to lug them up eight flights of stairs to some design office then end up having to swap their other screens out as well because the screens were different shades.

We have very few Sun keyboards and VDUs here, as they are extras.  99.9% of our stuff is rack mounted with a serial connection up its jacksie (or LAN to NetMgmt port if its from this millenium )

You've just reminded me, I haven't patched the newest TS's ILO and configured the port into the ILO Vlan yet... ah well, what can possibly go wrong. 

[TheBoy]

6.8 years 

Thats bloody impressive! 

Dunno how long my home server has been running since last reboot, will check and see in a moment. However it has been turned on 24/7 (minus the few hrs when its been off for hardware changes / UPS upgrades etc) for almost 5 years. Its only a crappy IBM eServer too! x206 IIRC... Which reminds me, need to shove a new backup tape in - its VERY VERY overdue 

I know none of the servers at home have long uptimes - the hypervisor that they all sit on has only been up approx 166 days.  Which is probably around the time when I swapped out the server hardware (July iirc)

[TheBoy]

You've just reminded me, I haven't patched the newest TS's ILO and configured the port into the ILO Vlan yet... ah well, what can possibly go wrong. 

iLO (as in HP) or ILOM (Sun).

HP iLOs tend to flash OK, although if its a B-class blade, if it fails, it needs BOTH motherboards replaced.  Sun ILOMs, well, I have a bit of a midas touch with them.  On SunFire T2000, I'm currently around a 30% success rate