RIP
That was about the last remnant of the old HP. The HP that did things well.
The final version of HPE's own flavor of Unix, HP-UX 11i v3, is now out of support. It is the end of a line that started in 1982. According to HPE's HP-UX support matrix, the end of life for the final version was the last day of last year: HP-UX 11i v3 11.31 HPE Integrity: standard support through 31-Dec-2025 The product …
Back in 2004 I was at a company that used HPUX Integrity Itanium servers for big Oracle databases. I recall they had redundant out of band management cards on them. At one point we did a big network upgrade moving from older 24-port 3COM switches(I don't remember the model but have a picture taken a year or two later http://elreg.nateamsden.com/2004-network.jpg ), that may of been gig? But I think they were still 10/100 on our backend network. The network was quite a mess, none of the cables were labeled and everything was going to a single rack, very tightly packed. We were going to a new pair of 48-port Extreme Networks Summit 400 switches.
The migration was basically unplug everything, remove the switches, put the new switches in again, and plug everything back in again. During this process we caused a network loop, somehow plugging both ends of the same cable into the switches, no idea how that happened but it took a while to track down just unplugging ports till the loop stopped. I recall at one point our manager was saying "we need to prepare to roll back soon". The network engineer and I(I was the server/app/ops guy) looked at each other and said, no way we can't roll back it will take so much longer.. anyway we found the loop and fixed it.
Where Itanium comes into this story is that loop caused all of the Itanium out of band management ports to hang, from what I recall there was no other systems that had lasting effects, once the loop was gone everything was fine except those out of band Itaniums. The back end engineer in charge of those systems contacted HP for suggestions. There may of been a reset button on the cards, if so, they tried it and it didn't help. HP insisted those cards were hot swappable, and to just yank the card and re-seat it and it would reboot with no impact to the running HPUX system.
So they yanked it on a lower priority system after hours and the system promptly crashed hard. It came back up, but clearly not hot swappable.
The next year or so the company funded a project of mine to basically rebuild everything, all new servers, new network, new back end storage, and new Itaniums (I had nothing to do with the back end). Built it out much better then, though we still made mistakes since we were learning as we went.
The company later moved to RHEL Oracle systems after I left on x86(when I first joined they ran HPUX on PA-RISC). I recall years later still being told their largest customer, AT&T was demanding they return to big iron Unix(didn't care which), and AT&T said they'd even foot the bill. Apparently Oracle on Linux was super unstable, at least for the databases they were running (largest single instance OLTP in the world, by that point maybe it was 70-ish TB, it was mid 50s TB when I left 2-3 years earlier).
HP used to be a good company.
Nowdays it has splintered into a handful (or more? I've lost track) of fair to middling companies, barely or not at all related.
I have friends working at a company which HPE bought; this is apparently not uncommon, as HPE's business model (so to speak) seems to be plucking things off the refuse pile in hopes of finding diamonds, but usually coming up empty, or near enough. I never got used to saying/typing "HPE" -- at least Agilent had the sense to differentiate themselves -- so I just carried on referring to their original company name.
It doesn't stop with the names, though; I'd be hard-pressed these days to tell you what HP does (printers, still, I guess?) vs. what HPE does, without looking it up.
Strange, but apparently true. After splitting into HP and Agilent (in 1999/2000, long before the HP/HPE split in 2015), Agilent spun off various parts, including its semiconductor components product group which became Avago, which evolved into Broadcom! There are very few very big corporations which survive indefinitely without major change, or collapse, or fragmentation.
Don't know but I still have my hp 16C calculator that I bought around 1977 and it is still in perfect working order. From the distant past when HP (AKA High Priced) made top of the line quality products at top of the line prices. Given that I once bought a TI scientific calculator who's keyboard quit working in less than a year I'd say that quite often quality is worth the price.
> hp 16C calculator that I bought around 1977 and it is still in perfect working order. ...when HP (AKA High Priced) made top of the line quality products at top of the line prices. ...I once bought a TI scientific calculator who's keyboard quit working in less than a year...
OTOH: I been rocking TI's TI-30 types since ~~1980 when they were still LED. Those still work fine but eat batteries. Since then I've always had two (work and home) TI TI-30Xa by my side. Yeah they are cheap knock-outs but they sell new for $12 and used 3 for $10, even in very rural flea markets (they were school-approved for many years). When one gives me trouble, even just faded battery, or illegible LCD (they used several part numbers), I grab another. Not counting my first, I have not spent $100 yet.
And not counting two slide-rules that I still use (some problems are easier on a slipstick).
Don’t they still make some decent calculators or have they died a death too?
I purchased a RPN HP35s perhaps +15 years ago was definitely "Hecho en China." Still going strong so fairly decent build quality.
According to wikip. production was discontinued in 2021 but the calculator store still lists them at €45 ex. + shipping.
Probably easier these days to install a HP calculator emulator on your phone — just about every model is available either gratis or for a few dollars.
I have Free42 (HP 42s emulator) on mine as it has similar layout to the 35s but mostly because I am cheap.†
† more a matter of not sharing any financial details whatsoever with Apple or Alphabet.
The calculators have all gone down the outsourcing route.
The HP 49G was the last "real" HP calculator. It was a great design (in terms of the firmware) but the key layout and feel were not popular.
After the 49G production was outsourced to Taiwan. The HP custom CPU was no longer used. Instead an ARM CPU emulated the older HP CPU.
> And then along came Failorina.
True story: I was at the VIP afterparty after LinuxCon 2016 in Dublin.
People noticed my badge...
A prominent Cloud consultant:
"Oh God, El Reg are here? Are you gong to write about our dress sense and call us eggheads or something?"
But the bit that amused me was when an HP staffer showed their deep engagement with the CEO:
"Ooh no, the big spilt is going to be great. We'll be great, we'll both going to do great. And our bit gets to keep Carly!"
No, they didn't. The _other_ bit got Carly. Much luck it brought them.
The party in 2016 must have been a good one.
Carly Fiorina left HP in 2005 having been recruited from outside the company to be CEO of HP when Lew Platt retired and Agilent (the non-computer parts of HP at the time) was spun off. Of course she carried out the HP/Compaq merger.
Carly Fiorina was succeeded by Mark Hurd, from NCR, whose notable contributions included the acquisition of EDS (which was followed by an $8bn write-down I think)
In 2010 Hurd was replaced and the disastrous appointment of Leo Apotheker (formerly of SAP) was made - he instigated the Autonomy purchase (even more than £8bn written down on that!). Meg Whitman (who had been at eBay amongst other roles) was already on the board at the time and she took over as CEO when Apotheker was ejected later in 2011.
Meg Whitman (definitely not Carly Fiorina) stayed until 2017 and in that time I think EDS was spun-out to DXC, HP's software businesses to Microfocus, and then the major split between HP (printers/PCs) and HPE (all the rest, e.g. storage, networking, servers) took place, with Meg Whitman heading up HPE until the end of 2017 when Antonio Neri (a long-term HP employee) became CEO.
I gave away an LJ4 a couple decades ago, I wasn't the original owner either, it was old when I got it.
It wouldn't surprise me at all if the thing is still out there somewhere, dutifully cranking out pages.
Though the cost of ink consumables was getting prohibitive by the time I passed it along, I can't imagine what they might be now, if you can even still find them.
Used to deploy a lot HP LJ 4050's work group laser printers back in the early 2000's and even those newer one's were workhorses. I saw a 4050DTN on the kerb outside a Dr's office a couple years ago and I nearly pulled over to rescue it becasue it was probably still working and only would have been replaced becasue it was slow compared to new laser printers or copier MFC's which everybody moved to. The nostalgia was strong and I only didn't do it becasue I had nowhere to put it in my home office..
It was a nice system, for a Unix, though the quality and general usefulness of the annual releases had been going down for a while - toward the end, they tended to be not much more than JVM patches and new (but still old) OpenSSL releases.
Quite a few Integrity customers are still on it, pending migration projects (almost entirely to Linux, but a few to AIX.) I had heard some hopes a year or two ago that maybe HPE would extend the EOL date, but seemingly no luck.
I'd rate HP-UX above AIX, though the latter gets a few points back due to smitty. HP's "SAM" tool was... okay, but that's all.
From an end-user perspective HP-UX was decent, close enough to Solaris with enough things that people coming from there didn't feel completely out of water.
But as a sysadmin who had to install, configure, and support most of the usual suspects 90's era Unixes for one particular $JOB, and then also build a whole suite of the common opensrc tools at the time on them, AIX and HP-UX were usually fighting for the bottom rung in my book. I (thankfully) have forgotten now, but it seemed like AIX had more trouble with GNU compiler tools than HP-UX, so it was harder to bootstrap that environment without outside help.
"it was weird and it required more hacks than aix, Solaris, or irix."
The HPUX Software Porting and Archive Depot (from memory) was your friend if you wanted gnu software — gcc for starters — but a fair bit of other open source software too.
For my sins I worked for roughly a decade in an environment where I was responsible for workstations running a variety of versions of decosf/1 (digital unix, tru64), hpux, irix, sunos (solaris) and aix. The only obvious commonality was they all supported NIS (YP) and NFS. Every version supported at least two file systems and no two bsd ufs derived file systems were compatible.
Some of the oddness was due to the OSF — UNIX International split. DECOSF and HPUX10 had a fair bit in common, while Solaris and Irix shared some commonality.I think only Solaris2 was a system V.4 derivative; hpux10, irix were System V.3; Aix was System III?; DECOSF had a Mach foundation but looked like System V + BSD (which could be said of most Unices.)
Still open source software was rather portable between various Unices - unheard of in other OS. I had ssh v1 running on all the boxen long before any vendor support . When much later postfix appeared, sendmail was quickly consigned to history. ;)
One thing I remember was once you got down to the command line utilities (required for scripting) a lot of the differences were mostly cosmetic.
In the event over a period of two years all the Unix boxen were replaced by inexpensive, faster x86 Linux systems. I still have collection of OS and software cdroms for hardware now only found in a museum or landfill.
I'll raise a glass to any of the old Unixes from way back when, because those were good times -- I felt like I was at the top of my craft as a sysadmin back then, looking after the metal and the OS, proprietary though it may have been.
But I can't go so far as saying I'll miss HP-UX (or AIX). I didn't much like working on either.
My fondness-nostalgia for HP-UX extends as far as it being the first job where I was paid to be root, on an HP 9000/500 series, where I also learned to thread reel-to-reel tapes for backups, initial forays into making serial cables (for all the Wy50, ACT5, and real DEC or other VT* clones in the office), plus a bit of very amateurish TeX for docs on occasion. Arguably it's where my sysadmin career started, so cheers for that. But years later, being forced/paid to touch HP-UX again on a C3600(?) hppa system and others, I can't say I enjoyed the return visit.
> It was a nice system, for a Unix
I was not able to find a categorical citation for this, but I am told that HP-UX was the first Unix with built-in containers: HP "Virtual Vaults" appeared in HP-UX 10.24 or so, roughly 1996.
https://bitsavers.org/pdf/hp/9000_hpux/HPUX_Versions.txt
This means before Solaris Zones (2005), and before FreeBSD Jails (2000).
No config files in /sbin/ Mostly, they were under /etc.
Are you thinking of the start-up scripts under /sbin/init.d ? Some people would hard-code configurations into the scripts in there, but it wasn't the "correct" way to do it. You were supposed to put the configuration changes in /etc/rc.config.d and the scripts in /sbin/init.d would read them.
JFS in HP-UX was basically a port of VxFS - the Vertitas Filesystem. It had nothing to do with IBM JFS.
The Logical Volume Managers on HP-UX and AIX were very similar conceptually, but I think they had entirely separate code bases.
(qualification - managed HP-UX systems for over 5 years and then sold HP-UX systems at HP for 20 years)
Based on what I recollect once seeing in an ancient SAP R/3 installation guide, both HP and Digital acquired LVM via the OSF entity of which they were members for a while, along with IBM (who contributed LVM of course). Either HP or Digital, or both, reworked LVM a bit, and created a new set of commands replacing those used by IBM (e.g. IBM "mklv" vs HP/DEC "lvcreate). HP-UX had added LVM from version 9.x onwards (it was hellish before that, when working with large disc configurations) and I assume it was Digital Tru64 (or OSF/1 predecessor?) that had it, rather than Ultrix. Digital gave up on LVM pretty quickly (moving to AdvFS?) but HP and IBM stuck it out, developing their own versions in different ways. The HP version (or at least the command set) made it across in some way to Linux as well, eventually.
> what HP did to Tru64
Tru64 was DEC, not HP.
Compaq acquired DEC, of course, and didn't really know what to do with it, as Compaq was a PC clone maker. At one point, perhaps _the_ PC clone maker, but even so: it didn't have a clue about big enterprise systems.
Then HP acquired Compaq... and proceeded to kill off most of the stuff that competed in any way with the offerings of HP's big BFF MICROS~1.
VMS was lucky to make it out alive. Not much else did. But HP had bet the farm on Itanium by then, so Intel ended up owning Alpha, and killed that, because of course Intel was BFFs with MS too.
It bet the farm, it bought the farm.
"Hewlett CompaqArd", as we called it for a little while.
Some of those Compaq-badged Alphas were genuine beasts, and some were also pretty nice to work on/in (depending on the cabinet size ;-) ), but perhaps my favorite DEC was an old tan AS1000 deskside with the original [digital] badge I rescued, and installed NetBSD.
DS10 and DS20 were much more practical, but that AS1000 was fun.
By the time Compaq acquired Digital it had already also acquired Tandem and was, I think, the leading vendor of Intel-based servers (starting with the Systempro which evolved into Proliant). That's more than being a "PC clone maker" which was also then true of Dell (who had the PowerEdge line then, as well as a wide range of PCs). "Big Enterprise Systems" were effectively doomed as soon as the Intel servers began to grow in capability.
For completeness, I'd say that one reason that VMS is one of the last "proprietary OS" men standing is that Compaq decided to port it to Itanium (which was logical, considering that they had decided not to develop Alpha further) and HP allowed that project to reach fruition.
Although Compaq had done some work on a Tru64 port to Itanium it could not have made sense to sell and support two different proprietary Unix variants on Itanium based servers once HP and Compaq had merged (the last Alphaservers were branded HP though). HP management didn't simply favour the HP legacy products - the HP3000/MPE environment remained stuck on PA-RISC so ended support long before VMS did, HP's Netserver range was replaced by Proliant , and HP VA storage was retired with DEC/Compaq EVA prevailing.
it means "if you call us and your problem turns out to be an existing bug we'll help you and give you the fix; if it turns out to be a new bug, we'll try and figure out a workaround, but no programmer is going to produce a patch to fix the bug"
Some orgs think that is worth paying for, some don't Fairly common with mature software across all sorts of IT companies.
Exactly - no new patches, including no new security patches. Here's what HPE actually say:
https://www.hpe.com/us/en/collaterals/collateral.4aa2-5741enw.html
Specifically look at the section entitled "Service limitations" which states the following:
"HPE Prior Software Version Support and HPE Mature Software Product Support services without sustaining engineering are limited to the resources available in HPE’s global service delivery organization. Problems, including security vulnerability, that cannot be resolved by the service delivery organization through providing advice or delivery of a commercially available software patch will be addressed via a workaround, where HPE can make such a workaround available. Problems will not be escalated to an HPE or third-party engineering group. No official security vulnerabilities status or fix shall be published."
The point being, there won't actually *be* a HP-UX engineering team to escalate a bug to, as all those engineers/developers will have been let go or reassigned. In addition, for the parts of HP-UX that were 3rd party (mainly the Veritas filesystem, but there are also performance monitors/agents that were integrated into the OS that are now owned by OpenText, who acquired Microfocus, who bought them from HP), any contracts/agreements to support/maintain code will now have lapsed.
I was a big fan of HP-UX back in the day as a rock-solid OS for big iron server systems (a lot of the criticism in the comments here comes from folks who mainly used it on workstations and/or as a developer platform where it was always inferior to Solaris), but anyone using it now is exposing themselves to an ever-growing security issue as new vulnerabilities will remain unpatched.
I'd say it now belongs in a museum, but what's left of HPE won't even do the right thing there and provide a hobbyist release as they did for VMS. Likely it will wither away to be completely unknown once those of us who used it are gone.
I did debugging of some stuff on PA-RISC I think. It was a weird architecture. Keep in mind, 32 bits, 4G of addressable memory. This was a big P&R program that needed it. But HP chopped up memory into chunks. Code got the 1st G, heap the 2nd, shared libs the 3rd and stack the 4th. Or something like that. It was not good as we wanted more for heap, and less for stack, and way less for shared libs. Sun if I remember right did the expected, code, heap, stack down from the top. And you try not to have them pass each other in the middle. The code was based on sun, so sometimes intractable problems on hp due to memory organization. Now of course, 64 bits. Bigger issue is how much memory can you afford.
"But HP chopped up memory into chunks." — called quadrants. A perl based front end to [k]adb was called "q4" presumably from that."
Some of the gory details are covered in hp-ux 11i internals (1st edition) Chris Cooper & Chris G. Moore (Prentice Hall 2004) which is an interesting read two decades later when both the architecture, PA-RISC 2.0, and finally the OS, HP·UX11, have been consigned to history.
"Support without Sustaining Engineering through at least 31-Dec-2028" is rather intimidating when I consider I am also on the same conveyor belt of obsolescence.
I always found AIX to be one of the better unixes. Early releases were a bit wierd as IBM marketing wanted features to distimguish it from other NIXs. Then they actually listened to thier customers who wanted a standard unix and went for 100% POSIX compliance. Still going strong and the hatdware ia ace.
This was probably the first and last time they took feedback form custpmers on board.
The first training course I did at my first job out of university was the HPUX 9 to 10 upgrade course. All about what was new and changed in the latest version. I mostly remember it for the change from the BSD init system to the SysV one. All those files with the crazy S1... K1... names seemed overly complex at the time! :-)
I remember we didn't use SAM (the admin utility) as there had been a bug in it (before my time) and we didn't quite trust it any more. So manual edits to /etc/passwd to create a new user or change their group, etc.
"my first job out of university" ... me too.
At the university, in 1993, I used SUN workstations. Great stuff
Then, in 1994: military service. No computers, just filling out paper forms
In 1995, I started working: Windows 3.11 on the desktop. Brrrrrrrrrrrr: 1 - 3 OS lockups/crashes per day was usual. Luckily enough there was a HPUX system I could use for the real stuff that needed to be automatic and reliable. And HP support was serious and good.
Ah, another trip down memory lane. Back (IIRC in the early noughties) my employer had acquired a retail poeration and we went out to buy a new EPOS system. The back end ran on HU-UX which was ... "a bit different" to the SCO OpenServer I'd learned on. Unfortunately, I didn't get much time to learn anything on it, as the Christmas sales period was "bad" and the administrators were called in (for the retail business which was legally separate) by the end of the holiday.
Yes I have a beard, yes it (and the rest of my hair) is grey.
Yes, in 1993 our usually deployed to SCO Unix application had to be ported to HP-UX v8. On the earliest of PA-RISC servers (an E class @40Mhz IIRC) .
It had a few weird quirks like a pre-partitioned disk (no LVM yet) but mostly worked fine. I then worked on upgrades thru 9, 10 and 11, on systems all over Europe. the middle of these projects - right up to 2023/2024 when some of the last of the old Itanium Superdomes was decommissioned.
I suspect that there may be parts of UK government especially defence which are still running HP-UX because they cannot agree what to replace it with. The tech is now changing too fast for their business.
I would like to pay tribute to the HP hardware engineers over those 31 years, who endured appalling cutbacks to their ability to keep the customers running.
Also the old HP-UX forums pre-revamp from 1998-2006 was peak internet as far as I'm concerned. I was Roylty level, then made honoured founder which sounded a bit too DS9 for my liking.
So many stories to tell, such as the K class which did a full 360 end over end down a flight of stairs and still rebooted with only 1 disk broken. This was solid hardware.
Agree - the ITRC forums were the best that you could ever want from community technical support forums. Lots of very bright people willing to help each other and a distinct lack of snark - people were very rarely told to RTFM, and even when they were, they were pointed at the bit of the FM they needed to read. Even better, there were one or two of the old HP-UX development lab folks who regularly contributed.
I made it all the way to "Olympian" level, as it was called - stayed on for a while when they revamped it, but it was never the same. Ended up losing my account when I left HPE - the hopeless support staff running the support centre by then (2023) couldn't figure out how to extricate it from my no-longer-valid HPE email address.