Sysadmin misses out on paycheck after student test runs amok Welcome once more to the world of Who, Me? – El Reg's weekly trip down memory lane with our dear readers. This week, our confessional column is taking a trip back to a college in the late ‘80s, with “Bryan”. While Bryan was at college, he was simulating a communication channel for binary data, which he described thus: “Take a …
did a student have access to the admin system in the first place?
Back in the day when I worked at the local college we'd certainly have noticed the excessive CPU times almost immediately as we reported on the heaviest users and applications weekly*.
*By far the biggest application for page swaps was a simple game which the students must have been playing endlessly. It had been written by one of the programmer/analysts who'd subsequently left but I found that by moving a subroutine from a separate function to inside the main code it would run without swapping. The system performance noticeably improved.
I agree with your first statement.
But it sounds like he only used the machine once and that happened to be the night the payroll was run... In which case, unless you were running Nagios or something and set it up to create an alarm for excessive CPU usage, you wouldn't find out until after the fact - depending on whether you had night shift operators or not; if there were night shift ops, they probably should have noticed something wasn't right.
@GlenP, yes, he had run the program several times on other machines and he had permission to do this at night, but it sounds like he chose the machines randomly, based on name and happened to pick Ada at a critical juncture.
@ anothercynic, And yes, I know Nagios wasn't around then, I was pointing out how we could do it now. Back then, as said, unless there were night ops keeping an eye out for "rogue" processes on the machine, you wouldn't have known, until after the fact.
Edit: On rereading, I guess I didn't make that clear, my bad.
No, he used it several times... but happened to assign a job to Ada the night payroll was running.
And you are talking about Nagios. You do realise that Nagios' predecessor was first released in 1996... Look at what the article says:
This week, our confessional column is taking a trip back to a college in the late ‘80s, with “Bryan”.
Back then, Nagios was not even a glint in its creator's eye, since he was barely in secondary school then...
Just sayin'...
Agree very much that access to administration systems should have been better controlled; however so rusty that I cannot recall if it was possible to restrict access to sub-sets of members of a single VAXcluster with a single UAF. I suspect that something could and should have been implemented to keep user accounts to their authorised systems!
Brings to mind a situation at one of what is now a Russell Group University back in the mid-1970s when parts of the student body took it into their heads to occupy the administration block over some political matter (Overseas Student Fees, if I recall correctly). Anyway, there we were in the middle of the night in various parts of the admin area (based in what was then said to be the longest corridor in Europe) when I realised that a small bunch of hot-heads had broken into the administration computer suite and were about to cause significant physical damage to the disc packs they found in the room (some variety of mid-sized ICL 1900). As a computing mainstream student I could see the potential for many types of 'own goal' if they went any further. Thankfully, with a bit of 'persuasion' I managed to get them to leave things alone and cause no further damage.
Spent the rest of the night re-fighting the Wars of the Roses (Kingmaker-plus) in the Vice Challencor's office - while carefully not inhaling the various substances being passed around.
(based in what was then said to be the longest corridor in Europe)
I suspect there are a lot of those.
I heard the claim made of Newnham College (Cambridge) when I had occasion to navigate it. I thought it a slightly dodgy claim: it worked by aggregating a lot of sections that wouldn't necessarily all be considered as a single corridor. Anyway, not a place I can imagine as a venue for what you describe, even if someone hadn't already outed you elsewhere :)
based in what was then said to be the longest corridor in Europe
I suspect there are a lot of those.
In much the same way that I know of at least two universities (Exeter and York), where a library was constructed without taking into account the weight of the books, and so the entire building is gently sliding downhill. I'm not sure if this is more than just a rumour at either though.
I cannot recall if it was possible to restrict access to sub-sets of members of a single VAXcluster
So you set up the system "login" script (whatever the VMS equivalent of /etc/profile is) to print out "this system is used for university payroll / administration, authorized users only" which would have been sufficient to stop this guy from trying to use it. Better yet, have it check against a list of users authorized for that payroll system and log off those who are not. I'm sure there's probably a way around that, but we're not trying to keep hackers out, we're just trying to make sure people don't accidentally do something they shouldn't.
When I was at uni (back when dinosaurs roamed the earth), the admin mainframe was in a subterranean fortress-like administration building and the student timeshare/batch mainframe was on the ground floor of the research center on the opposite side of campus.
They were, needless to say, completely separate operations and in no way connected. IIRC, the admin system was IBM, the student system was CDC.
// fortress-like, because it was the 70s and there was an unpopular war on
At least it wasn't as bad as a place I worked at in the early 80's when one of the cleaners unplugged the PDP11 payroll computer half-way through its run, to plug in her vacuum cleaner. It took a couple of days to get that one sorted out. (There's a good reason why hospitals etc. have special sockets just for vacuum cleaners)
No longer allowing vaccuuming or buffing in the data center. hint, the bottom row of tapes on every rack were wiped by the buffer... it took about a month to discover this, and it was only discovered when the CE was in laaate at night performing maintenance when the janitor rolled the buffer onto the raised floor.
I had an issue with cleaners knocking a cable out of the back of one of the VAXs that connected to an LP27. I would come in on my shift and find the jobs stuck in the queue, and while troubleshooting the problem, before calling DEC for service, I found the loose cable. The problem was I wasn't around when the cleaners were there most of the time, and then one day we crossed paths and the problem was solved.
I worked at a place that had wonky mains. Turning on the floor buffer caused the lights to dim. They turned on 3 IBM servers and they had weird issues. first time they did not come on. To much current draw . Week latter they caused so much issues when the disk spun up they called IBM and demanded IBM came out and saw that they did not have a PDU cabinet that they fixed it. They told them that a proper licensed sparky would needed to fix the issue . Oh and your warranty is voided .
Worst was realising at around midnight that my first application test run of my pride and joy - selecting jurors from electoral roll - had a fundamental flaw... and would keep selecting jurors forever as the test (have I reached EOF and do I have 20 jurors) was too specific. I leave that exercise...
Jumped in my car, drove 40 miles to office, pounded on random windows for 20 minutes until someone heard... and was told by the scary night shift operator that she'd assumed I'd made a balls up after it asked for the tape the third time and had killed it, and that she'd "deal with you in the morning".
Phew... sort of.
We never let student or faculty jobs run at the same priority level as the SIS (student information system) or Accounting jobs. There were users in the physics department who's jobs would run for over a week because they ran at such a low priority that any (i hesitate to say 'real job') (ah!) university critical job, like payroll, grades, etc. always ran with a higher priority. That being said, from about 11:30 pm until 7:45 am, the physics department could keep all of the processors at 100% for weeks at a time.
That being said, from about 11:30 pm until 7:45 am, the physics department could keep all of the processors at 100% for weeks at a time.
They told you they were running fission/fusion simulations, but in reality they were rolling in from the bar and playing Trek all night.
Some manual SQL interventions needed to payroll as lots of timesheet data was missing due to a related technical mishap, and change of legal entities doing the paying...
After payroll day was told, 2 people did not get paid at all. Colleague said, "Only two didn't get paid? That's not an issue." I did suggest he may have a change of attitude if he was one of the two...
"After payroll day was told, 2 people did not get paid at all. Colleague said, "Only two didn't get paid? That's not an issue." I did suggest he may have a change of attitude if he was one of the two..."
LOL, upvoted, but in his defence, he probably meant that dealing manually with only two payroll failures isn't an issue the way 20-30 or more would be.
I've groaned. I know there are parts of the world where Peanuts is thought to be about a dog called Snoopy with some kids as supporting characters - but still. Charles Schulz would be rolling in his grave if he hadn't always had healthy distance towards his work and fame.
The star - or the would-be star - is the show's unique selling point, so you name it after them, and not a snappy title that the audience is still less likely to remember. In some cases, "The Name Of Leading Performer Show" also has a different setting and even cast each week, aside from that one constant element. "Hancock's Half Hour" for instance didn't particularly start like that but ended up that way.
It's also a case where the star plays a character with the same name, which could confuse a stupid person and worries me - although with "Hancock", the show tended to be about what the cast of the show got up to when they weren't performing the show... except that that was, in fact, the show. Even more confusing. But it did make sense for Mr Tony Hancock to appear as Mr Tony Hancock.
> Some ensemble or supporting characters sometimes just stand out and become the star, regardless of the creator's intentions:
True but you don't normally see people referring to "MASH" as "Hawkeye" or to "Happy Days" as "Fonzie", the equivalent of which ("Snoopy characters" rather than "Peanuts characters") happened in this article.
On the other hand, peanuts wasn't really a thing in a lot of countries. Snoopy is however the most iconic of the Peanuts "cast" and thus for international readers Snoopy characters might be more recognizable than Peanuts characters. I don't think I've ever even seen a full Peanuts cartoon for instance.
Charles Schulz would be rolling in his grave if he hadn't always had healthy distance towards his work and fame.
He also wouldn't be rolling in his grave because he absolutely loathed the name "Peanuts", which was arbitrarily assigned to his work by the syndication company.
I once got an emergency call from our Mainframe Computing centre to say that the system they used for processing Prison Officer's overtime payments and gone haywire and I needed to get there NOW to fix it, or the POs would work to rule and this could lead to prison riot!
The system ran on an old CCP/M machine which each prison connected to over a dial-up link and transferred their data, which was then uploaded in to the mainframe to feed through to the payroll system.
I asked what was wrong with it, to be told that the on screen menus had all disappeared and they couldn't use it.
I jumped into a taxi to take me the 5 miles to the data centre, then waited for about 15 minutes to get let in. I then walked over to the machine, which had an old green-screen monitor, inspected it closely for about 30 seconds, turned up the contrast on the monitor and walked back out.
Someone had cleaned the machine, using wiped on the keyboard and monitor, and had managed to turn the rotary contrast knob right down, meaning thet the menu text and background merged in to each other.
Related: Back in the 1990's I got sent to Brutish Rail's Bletchley depo to fix a dead computer in the maintenance area. Got there, wiped rag over monitor that was covered in about 1/2" of dust, looked down at the now visible glowing instructions and said "sorted".
Got a nice cup of tea out it and everything.
Just last weekend I was sent over an hour away to a shop that had lost network connectivity to one of its cash register servers. The machine was up, but there was no link light. Got out the cable tester; it showed open circuit on all pairs. I poked around and discovered a patch cord plug that wasn't fully seated in its jack. Shoved it that extra millimeter home, the latch caught with a snap, and I was a genius.
I was told that it had been installed 2 weeks previously, and had been working fine. Just took that long to creep out far enough to lose connectivity.
Had to buy my own cuppa, but hey -- emergency overtime!
"and had managed to turn the rotary contrast knob right down"
yup , ive seen that .
plus many other idiotic reasons for screaming THE SYSTEM IS DOWN!!
the screen is off
the app is minimised
someone has opened ebay.com on top of it
someone has changed to a different screen in the app and user dosent recognise
user cant navigate to a particular part of the system.
and this is , of course , always the users opinion of their client PC not "the system"
You could start a holy war on that subject.
My place now has a generic naming scheme for servers in data centres and VM clusters, but back in the olden days the various teams picked different naming schemes for their kit, and the general rule was "If you don't recognise the name, you leave it well alone".
"Ada" is a bit generic, and could have been relevant to more than one naming scheme. That's just asking for your kit to be accidentally rebooted, or prematurely scrapped, especially if there's more than one of them and people have been lax with record-keeping.
At one point, I was the "go-to" person to suggest names for new machines if people were stuck for ideas. I did get told off for naming one machine"Sicknote", but the Occupational Health folks were pretty cool about it .
I once had the task of decommissioning any server in the estate that didn't run to its owner when I came swinging my axe. After sending a seemingly endless stream of file servers called Gandalf to a well-earned rest, I was relieved to find a development team that had named theirs after the Second Stage Lensmen.
"I once had the task of decommissioning any server in the estate that didn't run to its owner when I came swinging my axe."
Ok, I assume that by "estate" you don't mean a big house with a big lawn.
What, exactly, does the term mean... various mentions seem to mean various things in various sources, none of them, probably, with big lawns.
Working for NHS Direct about 15 years ago, the convention for clusters was 3 letters of data centre name, followed by a hyphen the the name of a planet. Our DC was in Wakefield, so WAK- but for some reason the senior management objected to following Jupiter and Saturn with the 3rd largest planet...
In the late 80's we had a rather militaristic SysOp who liked naming his VAX systems and even the major storage media clusters after various battle gods such as Thor, Odin, Zeus, Mars, etc, etc. AND if you didn't get the message of NOT to mess with "HIS" machines, the VERY REAL and VERY WELL SHARPENED full-working-replica Battle-Axe and Broad-Sword hanging above his desk with an inscription of "Situation All F***** Up? - Beheadings Normal!" ENSURED that people got the point ESPECIALLY since the Sysop was 6'5" (2+ metres tall!) and probably about 300 lbs of Like-The-Rock musculature...which is the COMPLETE OPPOSITE of what many think a SysOp SHOULD LOOK LIKE!!!
I should note his systems ALWAYS worked like a charm and that there actually WERE FULLY TESTED Disaster Recovery Systems and MULTIPLE FULLY-TESTED ON A WEEKLY BASIS On-Site and Off-Site Full Image and Incremental backups "READY FOR ACTION" at a moment's notice!
Uptime on "HIS" systems was literally calculated at Six-Sigma (i.e. 99.9999% per year) uptime before that term even became popular in industry! We had people from major corporations and worldwide NGO and Government organizations come in all the time for VERY EXPENSIVE courses (5 figure sums for EACH student!) on how to keep six sigma uptime on major VAX and IBM computer systems like ours! I worked there only a few years (3) but even I must admit that was a VERY VERY WELL-RUN compute centre with great IT security and fantastic hardware/software recovery systems in place!
I do ALL the procedures then espoused and learned all those years ago IN FULL NOW on our own much-larger-in-scale systems! And YES it's Six-Sigma uptime !!!!
OK Zorac - I'm stumped on that unless you meant Zen (One of my favorites) & Orac, along with The System & The Alta (Granted the last one was applied to mainly to System augmented meat sacks).
"The Altas? They're linked into the System. They're not really people at all".
Edit: ZORAC, the shipboard computer aboard the ancient spacecraft in The Gentle Giants of Ganymede and the related series by James P. Hogan (1978).
Atlas, as in Charles Atlas, you know the guy who could hold the world on his shoulders? Just kidding, it's this guy:
Atlas was a legendary Titan. He was in a lot of stories, starting with a story including a hero named Hercules. Atlas led a battle of Titans against Zeus which lasted for ten years. Eventually Atlas was singled out by Zeus and forced to hold up the heavens as a special punishment.
We have a group of three compute servers (64-core Opteron machines) that three groups ordered together. The fluid dynamics group's one is called Poseidon, the AI group's one is called Pallas, and mine is dubbed Zeus, king of gods, as it is the most expensive one, with 512 GB instead of 256 GB RAM. Fairly boring, compared to some of the schemes mentioned.
I also rather liked the names of various Beowulf clusters in the past, like Hrothgar, and other figures from that epic poem.I suggested calling a cluster at our university Grendel, but that didn't take off, for some reason.
" is the most expensive one, with 512 GB instead of 256 GB RAM"
When we had a colourful naming system the new fancy schmancy machine that arrived and started the convention was HAL, first in a line of psychotic-robot server names.
I remember the server tech proudly proclaiming "Its got half a gig of RAM...."
If your machines got half a terabyte, it surely its new enough to be on the newer 21st century, widely adopted, no bullshit naming system of either a) whatever its asset tag is , or b) some sort of code with numbers e.g. SRV_WEB_007
Another one! Using underscores in server names; as a DNS admin I'm permitted to eviscerate you slowly for that.
Personally I really rather hate hostnames with numbers on the end ("server1", "server2", etc.). One of the key features of a naming scheme should be to avoid typos generating valid hostnames.
One UK university I worked at had servers named after colours red, blue .. white, black. Then sable followed black. Then ermine followed sable. Then weasel, and possibly even wolverine. God knows where it went after that...
My own research cluster of five HP-UX workstations had gordon (the big one), and then henry, james, edward and thomas
When I worked in ICL, we once mananaged a group of machines named after various poisions - cyanide, arsenic etc.
That's not the story.
Back then, the internal networks were OSLAN not IP - basically bridged ethernet rather than routed. This meant that (when auto-discovery/learning wasn't available), paths had to be created manually across the country to allow a client and server to see each other on the same apparent lan.
Because of this, many routes didn't work, so most sites had various login servers that could connect just about anywhere. - You'd connect to this machine, login as the host you required, and be connected across.
Someone automated the setting up of most of these accounts, and set up the gecos (user information) field to say something like "Remote login to XXXXXXX" (it was longer than that, but I forget exactly)
The point is, the "finger" command truncated this field at a certain point, so on many machines, if you typed "finger" you'd see something like:
LOGIN NAME TTY UDLE ....
arsenic Remote login to arse cuaa0 01
arsenic Remote login to arse cuaa1 01
arsenic Remote login to arse cuaa2 06
Yes, I never grew up...
I named our department after Ibizan nightclubs... I never told the bosses that though... Instead I came up with crappy fake explainations:
amnesia - that was the machine that was rebuilt requiring new memory.
space - that one had loads of free disk space
privilege - The new fast server that it was an honour to be using.
...
Alas no more, the names still exist on my mail and dns servers
My local university used a "Gandalf box" (the company name was Gandalf) back in the late 80s for serial connections back to various computers (HP2000, Pr1me, Encore, I'm sure other stuff too) There were a couple rotary selectors on the front that you'd set to the number of whatever computer you wanted to connect your terminal to. There used to be hundreds if not thousands of Gandalf boxes all over campus. Can't remember if they were 2400 or 9600 bps, but it was faster than you could read to be sure which was all that really mattered back when everything was text.
Once twisted pair ethernet became a thing, Gandalf went out of business and probably all those blue boxes ended up landfills.
OMG! The Gandaf Hybrid 2400 to 9600 baud modem with that weird advertisement in DEC-VAX oriented system admin magazines of a curly headed Centaur (man with legs and body of a horse) clicking together a giant multi-pin comms connector! That ad ran for YEARS and MANY people I know in the industry took those blue modems home after being retired and STILL use them as backup dialup connections if the main ADSL/Cable/Fibre lines go down! They STILL WORK after 30 years!
I still have five of them JUST IN CASE!!!! They STILL work and they were built TOUGH and even the rack-mount 2RU models I think I have one somewhere. If the fibre phone lines go down, I still have access to my 9600 to 56k DTMF-oriented modems which can be put out on a ShortWave radio for long range emergency digital communications amongst my fellow tech nerds who also have such gear!
Ibiza Clubs, that is a good one.
A certain popular and influential Magazine in the 90s (really good paper, fabulous ink) spun out a smaller new Digital version that had been growing within it for a while (they were tired of us). Naturally we inherited (borrowed, liberated, re-purposed, absconded with, nicked, jacked, and stole) many of their older SUN boxen for our grand little experiment. Out of respect we named our machines after ways of getting Wired.
Being good admins, out servers were Clearly Labeled (we also stole the label-maker) with the FQDN and it's internal IP and external (if any). Then we got to work making a new company and didn't think much of it...until The CEO™ (we had two; Ours, and the one from The Parent Company™, guess which one this was) decided to Take a Tour through our nice new shiny Server Room, to look at the blinking-effing-lights, at like 0100 hours.
Apparently some people are offended by servers named speed and cocaine and caffeine and espresso and some other things (also the NAT, but then aren't we all?). Come the next day, we were politely asked to change the naming convention of Our Servers in Our Server Room in an email.
Well, about this time, Fall Comdex had just finished up and we had just returned from Las Vegas (nice and hung-over and rowdy).
First order of business was the reply email saying it would be handled Oh-Fearless-Leader and so we renamed all the machines after Casino/Hotels on the Vegas Strip. [Point of the story.]
Second order of business was informing Facilities that if anyone from the C-Suite *ever* stepped in to the server room at 0100, with other like and well lubricated folk, unsupervised by a qualified entity, he'd be First on the Wall, *not* the suits (though they would get their turn), and security-card access was adjusted accordingly. [Moral of the story.]
Where I work, all the Windows and VMware servers have generic names, but the Linux servers all have names which, if you think sideways, are connected with what they are or what they do. Calamari - squid server. Postbox - internal SMTP relay. Termite - log reducer. Pigsty - has a lot of grunt. Stirfry is at our site referred to as WOK.
One place I worked at had every PC named after a muppet character. Some were pretty obscure.
Another place had their servers named after tennis stars of the 80's - Bjorg, McEnroe, etc. They were great servers (groan).
Then they were bought out by a big corporation, who insisted all servers were renamed to a boring convention of letters and numbers, and also insisted the "Caution: Respiratory Protection Required" warning sticker be removed from the door of the gents' toilets.
At my last job (at a high school), our server room was a former basement cafeteria. Complete with electric panel, water main/meter, and gas main, all entering the same room as the servers, but that's another story.
Being that it was in the basement, and a former cafeteria, there was a floor drain right under the 3 server racks. That wasn't a problem until a cleaning person unplugged the basement lifting station pump to plug their vacuum in. The sinks and toilets flowed into the sump pit, and when that was full, nature took its' course, and the "water" flowed out of the next lowest spot it could: underneath the servers. No raised floors here, just 3 server racks surrounded by an inch of poop water.
Needless to say, after the *third* time that happened, they plugged the floor drain. Not sure where it goes to now.
Indeed SID was the best actor in the series, like Orac in Blakes 7. But those costumes for the moonbase and skydiver Lady crew members were magnificent.
The box sets of UFO and Space 1999 are sitting next to my chair and competing with Solo for tonight’s viewing, quite a quandary.
Gabrielle Drake never wore a string vest - those were the uniforms for the Skydiver submarine and ALL the female operatives had a vest underneath. Purple wigs and silver mini-skirts were Moonbase uniforms. The future Mrs Michael Caine was also a Moonbase operative in a purple wig.
However, Gabrielle Drake in the movie Au Pair Girls... now you are talking.. .well drooling actually!
Our mainframe machines all have 4-character names, and most of them are comedians or cartoons. ERIC and ERNI and the stand-out ones, but there's also a TOAD and a RATY (if there was a MOLE, it's long been decommissioned). All the more modern machines have boring names with sensible schema that let you work out who they belong to and such by sight.
(AC because they're still in use, and probably ID my employer to anyone who's touched them).
"Several years ago" we installed PDP11/M micros in our offices in various towns in England. These were given three-letter names, usually the first three letters of the town name, like BRI for Brighton.
A new office was opened in Bristol, but the machine name was already taken.
The obvious three-letter name (containing two Ts) was vetoed by Management (boo, hiss), so it had to be called BST.
Back in the 90s I worked in a radio station that had a shiny new Siemens telephone system with a small LCD display on each of the phones. There wern't many characters to work with, so I had to be creative when programming up the system using the VT220 in the cupboard where the PABX lived - no full-screen editor, just a CLI and a PITA to use, if I may be forgiven some TLAs.
We hired two new DJs for the morning show, Mike and Steve. I set up their phones as 'S & M in the AM' and of course the best part is that they never got to see the text as it only appeared on the recipient's phones.
it was a full 10 days before the boss noticed and I was politely asked to think of something less imaginative.
(Screaming loudly, banging head.) You guys are giving me a PTSD flashback to when I did my MSc. back in the early 90's. We were required to use Ada for our final project and to take a course in Ada beforehand. We had '386 computers at home and work, and they all ran Windows 3.1. We also had a really funky Ada compiler which was all we students could afford to buy at the time. It had the wonderful ability to compile code that ran fine on the computer you compiled it on, but would not run on any other computer. Given that one of our instructors insisted that we turn in everything as an executable (as well as source code, of course), we had a wonderful time getting downgraded when the executables, which ran fine on our computers at home, failed to run on his. Somehow I managed to get an "A" out of him, but I never worked harder or had more stress in any class.
If anyone else out there is using Ada, I sure hope you have a better compiler than we did!
Should've kept this Ada away from students: tick. Keep her virtue intact.
Some comments here are evidently from the young, to whom that is and always has been elementary first principles. But it was a different world back then. People on computers and networks were interested, not malicious. Security as we know it today wasn't even a twinkle in the bursar's eye - let alone the sysop who was in the job for the exposure to otherwise-far-too-expensive toys.
"Nah, I tend to stick to boring, functional names."
Yeah, clever or geeky naming conventions only really work when you've only got up to maybe a dozen or so boxes to look after as was pretty usual back in the 80's or 90's. When it becomes dozen, 100's or even 1000's, you need a proper boring naming convention, especially when said server are all identical looking boxes in a server room, or even scattered around the globe in remote data centres.
If y'all want a decent naming convention, my friend down the road from me with his exa-scale systems uses a "Fun Name" (i.e. TEALC, CARTER, JACKSON, ONEILL, VALA, MCKAY, RONON, etc which are characters from the Stargate TV shows) and then appends a multi-digit 3D-XYZ coordinate.
Example:
TEALC-X122-Y271-Z1083-N127 indicating the network or cluster name, the actual physical cabinet number within the cluster, the shelf-number (i.e. RU rack case number counting up from the bottom of the 72 inch rack), and physical CPU number within each case. If we needed a physical core-number (some CPU's have up to 128 cores!) we would append an N-number at the end of the 3D-XYZ coordinate. This allowed our internal log systems to perfectly detail in physical real-world space where a problem piece of hardware is located.
It is ALSO useful for creating Virtual Processing Cluster where sub-jobs and sub-tasks could be assigned and tracked to specific locations within this HUGE data processing centre ensuring that no maintenance or downtime took place when those jobs were assigned to a specific geographic place.
Since we use a custom machine and networking OS that allows such long machine names, we can get away with such a naming convention. In Windows or Linux flavours, you assign each name as a DRIVE and SUB-FOLDER structure
Example:
Windows Server Name structure for all log files and data processing output.
\\TEALC\X122\Y271\Z014\N127
We use an X+four-digit-number for our coordinate names but you can make it 5 to 8 digits or more if you have GIGANTIC server systems with tends of thousands or even millions of devices such as:
\\TEALC\X81900122\Y02000271\Z01412345\N00000127
You can ALSO replace the X,Y,Z and N prefix monikers with C=Cabinet, S=Shelf, P=Processor, C=Core
so you get
\\ONEILL\C122\S271\P014\C127
which is fairly easy to read and learn where machines and sub-systems are in 3D-space.
At my last job (at a high school), our server room was a former basement cafeteria. Complete with electric panel, water main/meter, and gas main, all entering the same room as the servers, but that's another story.
Being that it was in the basement, and a former cafeteria, there was a floor drain right under the 3 server racks. That wasn't a problem until a cleaning person unplugged the basement lifting station pump to plug their vacuum in. The sinks and toilets flowed into the sump pit, and when that was full, nature took its' course, and the "water" flowed out of the next lowest spot it could: underneath the servers. No raised floors here, just 3 server racks surrounded by an inch of poop water.
Needless to say, after the *third* time that happened, they plugged the floor drain. Not sure where it goes to now.
I used to like systematic naming conventions until a major data centre replatforming project (several hundred applications and several thousand servers).
Changing hostnames was deemed too high risk (IP address changes broke enough things in poorly architected apps), so physical servers for one legacy DC moved to virtual servers in another DC retaining their old location based name.
Cue several years of telling people to ignore what they thought they could imply from the server's name when assessing change and to actually look at the detailed change plans.
Now I think that going back to random server names would be safer.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
