IBM Hardware Management Console
Вставка
- Опубліковано 26 чер 2024
- Setting up an IBM HMC, a Hardware Management Console to remotely manage an IBM System p5 POWER machine from 2006. The HMC is a period-correct IBM eServer xSeries 336 from 2006.
Check me out on Patreon: / clabretro
2.5" SSD Converter Tray: amzn.to/3N0hURt
Rack stuff
StarTech 25U Rack: amzn.to/3mEB7hS
Tripp Lite SMART1500LCD UPS: amzn.to/3KZW3Jw
1U 24 Port Patch Panel: amzn.to/3Nm0bFa
1U Brush Panel: amzn.to/3mExAA3
1U Rack Shelf: amzn.to/3oaDclT
Note: The above are Amazon affiliate links. It doesn't cost you extra, but I'll receive a commission which will help keep the content coming. I only link to things I've personally ordered.
Music by Karl Casey @ White Bat Audio
00:00 Intro
01:31 xSeries 336 to Hardware Overview
07:06 Getting the xSeries 336 to Boot
11:46 System p5 Hardware Fixes Overview
12:50 System p5 PSU Hot Swap
15:30 System p5 RAM
16:33 System p5 Hard Drives
18:21 Remote Serial Access for the p5
20:54 MRV Secure Console Overview
22:25 System p5 ASMI
22:57 Booting the System p5
24:19 Installing AIX on the System p5
26:54 SSD in the xSeries 336
29:00 Finishing the AIX Install
32:05 xSeries 336 Remote Access Overview
32:35 HMC 7.7.9 Software Install
36:54 Setting up the HMC
38:30 Exploring the HMC
39:48 LPAR Experiment
42:37 Outro - Наука та технологія
I like how the replacement PSU gets DC power from its twin in order to diagnose itself and blink its lights.
Really cool, Proliants can make it too!
The AIX menu system is called SMIT. The command line version is called smitty. The commands to navigate through the menus and list options are the bottoms of the screen. The idea behind it was they could allow for anybody to administrate and perform basic tasks with it. We builr up custom SMIT screens for our software and lab tools so that people could manage the system easily.
One of the nice features is that SMIT can show you the commands it is about to run by pressing the F6 button. As you showed at login, SMIT also logs all of the commands into a plain text log file so that you can see what was done under the covers. it's a great way to learn how AIX does things.
It's still my favorite UNIX.
yeah it actually wasn't too bad, once I got familiar. and agree it's really nice to see all the key options at the bottom
I started the computer club in my high school (circa 2010/11) and my school basically gave us free reign of any computer equipment that had been stored for longer than a decade. I took home old Pre LPO Tape storage libraries, big lead acid UPSs, that same GL380 you played with earlier. We upgraded our server to a Dell Poweredge R905 (I might be wrong but it's quite similar). If you ever see one you HAVE to pick it up. It had 4 P4 Era Xeon CPU's with 24GBs of ram. It used these crazy ram catties you installed in the front. I think it used 512MB DDR sticks. You had to install a weird kernel patch that somehow gave Windows 2003 32bit access to all of it. It took me hours to figure out how to do it when I was 14. They are so rare but out of all of the servers I've owned outside of blade servers, that thing was the most interesting. I have always wanted to play with one of these Power Servers though
interesting, I'll have to take a look. enterprise servers sporting pentium 3s or 4s are always interesting!
@@clabretro Proliant DL580 g4 is the HP from the same gen, it had the last and fastest netburst cpus, up to quad socket Xeon 7140Ms. 800mhz fsb, dual core hyperthreading clocked at 3.4GHZ with an insane 16mb l3 cache each. It was surprisingly fast in modern apps, at least the ones it could open (missing quite a few instruction sets vs a modern pc)
Nice machine. ~2005 is a period where designers tool full "advantage" of Java applets. Every server, switch, router and firewall required Internet Explorer, specific Java version with all security disabled.
Window manager of Hardware Management Console looks very similar to Fluxbox, or Blackbox. I recognize it by window decorator theme.
Lpars are still beyond kvm in Linux. This 50% of cpu doesn't even has to be a fixed core. Lpar will jump between cores and thread, depends of which one has a spare time for it. Not every feature will work with Linux VM's, they have to be AIX also.
I don't know about ram, but AIX is licensed per core. You totally can have more cores, which then are spares in case of CPU failure.
It was indeed Fluxbox, good eye. Yeah that LPAR CPU usage stuff is very interesting, I'll be playing around with it.
The last time I was dealing with a HMC was late 90's, early 2000's. We had a z390 mainframe with LPARs for the production, regular development, and a Y2K test environment. The HMC was a OS/2 Warp base machine.
Only touched these things as a DBA, so the exotic and interesting Power hardware was kept away from me and interacted only through a boring DB console. Great to see these being tinkered with at a low level.
interesting! and here I am not even running anything on it, ha.
I think those "light tubes" are actually known as "light pipes".
But man, it's so fun to live vicariously through your adventures with all of this old gear. I wish I had the room and the time to do similar things!
Ha yeah they might be, I keep using both terms interchangeably. Thanks for watching!
Ahh. Smarch 48th, 1980. I remember it well.
Lousy Smarch weather as usual.
😆
Quick read on the network card error, but I think what it was saying was that the card was actually "in the wrong PCI slot" so it was responding as intended.
I'm pretty sure one of those PCI-X slots is 100MHz and one is 133MHz. The card is intended for the 100MHz slot, but was installed in the 133MHz slot, meaning it was 'installed in the wrong slot'. If there were two cards installed, the system would assume that you intended for that to happen because something else was already in the other slot, but when there's only one card present, it's 'working as intended' to advise you that you could be using a 133MHz slot for a faster card.
On HP it was supporting all PCI standards, like these slots, Millennium GPU Card, NIC too. better than AGP !
Hey brother, I was thinking that serial IO card would of been used a PLC. I once hooked up a server in a cement quarry to with some fancy weight scales. Because it was that sensitive they needed to stick the server right in where the conveyor system was. The whole plant was running on a custom PLC was pretty damn cool to wire up!
Have to ROFL at your soldering and putting the battery in backwards, we're all human mate!
oh fascinating!
wow, took me over 2 years to realize FTP is not longer supported by default in browsers
Oh cool I used to work on these type of machines when I was working at IBM. I did hardware repairs and replacements. I don’t miss 3am data center visits haha
It still blows my mind that most of the hardware can be replaced while the system is running.
nice! yeah hot swap is so cool
Those x336 units were rock solid and super reliable. I ran these as load balancers with 4GB RAM and 2x 73GB drives until around 2014 when we threw (racks and racks of) them in the dumpster. Replaced with VMs that do the same job and boot in about 20 seconds :) Error 162 is time/date not set. Good to see the old P5 running.
This brings back memories... Years ago I set up a brand new HMC together with a brand new iSeries machine. There was a ton of IBM documentation to go with it so it wasn't too difficult. And the HMC looks very much the same as what you showed in this video. Besides two IBM i LPARs I also installed AIX in one. I remember I could move CPU's in units of one tenth between the LPARs.
I remember that at that time two LPARs could not share one DVD drives. So we had to order two in order to not have to go in to the HMC every time the other LPAR needed it. I think IBM has solved this problem by now.
I also remember that this iSeries model had the exact same three fans, and one of them died. The support contract had ended so I bought a second hand one and put that one in. A year later the other two died as well...
Nice video 🙂
Very cool. The LPARs are fascinating, I got some more RAM in so I'm hoping with my single two-core CPU I'll be able to setup at least two LPARs on this thing. Funny about the fans... I guess I'll count myself lucky so far!
Nice set of machines! One thing to note about the RAID controller on these pServers is in order to actually config the RAID array, you must use the "Standalone Diagnostics CD" to interact with the controller. Now as far as the whole Capacity On Demand thing goes (HW licensing), I believe it only pertains to dynamically adding or removing resources from LPARs, not 100% on that since my only source on that is some obscure sales brochure and my P6 not caring about me playing musical CPU books.
Other than that nice vid! I look forward to more POWER goodness in the future!
thanks! yeah as far as I could tell it wasn't able to do any capacity on demand stuff
On the System i machines of that era, you could in fact set up a limited number of LPARs without the HMC. (I think it was up to four) My vague recollection is that it involved booting into a special mode with the front panel keys to get to the configuration utility, so you couldn't modify them on the fly like you can with an HMC.
There was one big caveat, though. The system could never have been connected to an HMC! If it had ever been connected to an HMC, no LPARs for you without buying an HMC. If your VAR did some initial setup using an HMC before delivering the system you will not be going to space today.
interesting!
I wouldn't be surprised if that "dead" power supply is fine and just needs a fan.
Yeah I bet it's something simple!
You can on that era of system actually manage LPAR's via the IVM system (it's another service like ASM/ASMi) with PowerVM. No HMC required.. though going via the HMC I feel is always the better way and you can have much more serious setups with a lot of options for the VIOS that is ..just not fun over IVM. Very cool to see someone else hobbying this. Nice to know I'm not the only crazy out there. Though I work with this stuff, getting power5-7 systems working at home is glorious fun. :) Really looking forward to your LPAR and WPAR journey. Also those eServer disks that don't work maybe from AS/400 (aka iSeries). They use non-512byte disk block size because of how the tagged memory addressing works with the primary vs secondary storage concepts. Some are 540bytes for instance.. feeding those to a system that expects 512byte blocks will give it gas. If you've got some PC with an old adaptec SCSI card in it that has the "Low Level Format" util in it, you can specify a 512byte block size.. leave that tonking away for a day or so. Shake-stir-repeat for each drive. Also on that machine.. licensing also, you can have Linux/AIX/iOS (formerly OS/400 .. ) on the same system in different LPAR's. Back to OS/400 disks, if you're able to run iOS ..probably v5r3 or v6r$early .. you might want some of those disks on a controller that you can dedicate to the AS/400 LPAR and give it direct access to the storage adapter or have a VIOS that knows how to drive >512b sectors corectly. Best Wishes to you.
Thanks for all the info! Yeah I figured I needed to do some sort of low level formatting, unlikely something is actually wrong with the drives. When I dive into the LPAR stuff I'll probably throw dedicated network and SCSI controller boards in there to assign to the second LPAR to keep things sane. Totally new to it all, though, it's fascinating stuff.
You're a professional! And I love watching the professional at work!
I used to work on this back in the days. was fun to see it again.
Woo I'm early to a video for once! This era of IBM stuff is really great, I'm excited to see how this goes!
I love the older IBM machines but man do I hate the flexible hard drive cage with the cover that gets stuck because when changing drives I have had to rework some of those metal tabs when they get caught on the hard drive mount.
Ah.. Finally the HMC video 👍🏻 So when is the CSM video ? 😉 Yep another IBM Acronym 🤗 Before the CSM was CWS… Fun fact, the lowest CPU you can give an LPAR is 1/10th 😉
interesting! thought it was fascinating that you could assign partial CPU amounts
@@clabretro Yeah. These setups were/are used a lot in the Financial sector. The idea as you have already mentioned in the video, is to make LPARS from the physical racked system. If possible you can create something called a CSM (Formally called a CWS). This would house your MKSYSB images so that once you have created other LPARs on the rack you could then install AIX (or something else) automatically without going through the manual method you did in the video. So the 1/10th CPU and RAM you assigned would be good for that. As a side note. The HMC is a ideal solution for adding CPU and RAM on the go when an LPAR is running to help out any performance issues. 👍🏻
I love seeing the Linksys power stack in the background! I found this channel from that video, and I really enjoy your content.
Haha someone finally noticed. Thanks!
Great video! Was waiting for another appearance of this machine.
After cleaning up the M2, I totally thought you were gonna use that for the HMC server
We owned a bunch of IBM blade centers from that area, so the management interface looks really familiar. But the hardware was a real bad investment as the blade centers started to break down after a few years and our IT department wasn't able to keep more than two "enclosures" running. So from over 200 blades full of CPUs and memory we could only keep 28 alive and running and nobody wanted to use them for important stuff.
those IBM blade centers look cool, interesting to hear they weren't reliable!
For LPARs:
The way you booted the machine is referred to as "Full System Partition", meaning booting it without partitioning the resources... Which is fine for a P52A, but is hilarious when you realise you can do the same to a p595 system, which is two racks wide in it's full configuration! Booted, this way, the partition "name" is by default the serial number of the machine. IIRC that's always partition ID 1? but unsure rn.
On the Memory, AIX is finnicky with one bit, which is the size of some kernel-reserved memory within the OS is proportional to the next power-of-two of maximum memory allocated (because AIX can do hot add and remove, for both processor and memory! but, that may be a separate feature). so the best practice is to always set the "Maximum" value to the next "round" value, 1024Mb in your case.
On I/O, you can't pass individual disks to the machines, only full controllers (fractions of that would be handled by Virtual Adapters on a VIO, but you're not licenced for that). So you want to make sure different disks are on different controllers to start two of them up. Same for NICs, and because you only have one, you won't be able to get a second LPAR up in the network sadly.
One thing though, is AIX is really not picky at all about partition sizes and resources (you can run `cfgmgr` to rescan all hardware, if needed, even if they were hot-added). So, because you already have an AIX install, you could just delete the partition, create a new one with the controller that has that drive, and configure that partition's boot order to just boot off it, and.. it should just go, really!
Also if you do manage to get a second NIC, it'd be very interesting to see you set a Network Install Manager (NIM) in one partition, and have it do automated install on a second one!
Thank you! I actually have tons of extra NICs and SCSI controllers for this machine, so I'm thinking I'll add those in when I try to figure out how to configure LPARs correctly. Sad to hear I can't do the fractional disks... all of this makes me want to hunt down more p5 machines with the right licenses 😆
that violent network cable grabbing is giving me anxiety 😄
Wait till you discover live kernel updates and live swapping of instances between different metal.
Heh UA-cam knows my interests. Glad I found your channel!
My Homelab consists of an HPE DL20 G9, HPE DL380 G9, Unifi Dream Machine, Unifi Switch Enterprise 8 POE, Unifi Switch Aggregation, Unifi UAP 6 Enterprise, and some Tripp Lite power gear.
Playing around with the old hardware on your channel looks fun! The old hardware I play around with is mostly just PPC Macs but you're giving me ideas!
Nice! I've got a similar setup, except swap the HPs for a Dell R720 and a Dell R510.
I'm a professional
I'm dead that was so good, hahaha
only the best down here
direct thumbs up after the intro :)
Okay, tuning out the Thanksgiving family to watch. 😁
😂
Bet I am not the first one to tell you about an IODD ST300, I hope? I love these things and burning can be way more tedious than throwing an ISO onto a drive especially if you wanna try several versions. Man this IBM Stuff gives one nightmares with all the licensing stuff going on, and it still is a nightmare in the current world with IBM Mainfraimes. They sure do like to press you for the milk money...
It’s almost 4AM and here I am
hell yeah
It's 4 AM here too. ;-)
43min og pure goodies!!!🎉
This makes me want to build a business logo using light path tubes for my studio...
Ah man I gotta catch up. Was out of a computer for a week and a half. Mobo died and had to source a z590 for my workstation. Nabbed a killer deal on Amazon prime for it and a new air cooler that’s better than my AIO.
11900KF do be running hot, that 195TDP is a lie, I’m not even overclocked. But nailing 60c max on heavy synthetic loads.
Modular DB9 to RJ45 F/F adapters are your friend, they're cheap & easy to put together in whatever pinout you need. Put one together that handles the null modem then just use a regular network patch cable to connect back to your terminal server, bonus points if you color code those cables differently than the actual network patches.
Oh yeah totally agree, my plan is to get a bunch of those adapters wired up for all my use cases!
Glad I'm not the only one! 😅
Your eServer is missing a transparent plastic thing that goes where the other cpu socket is. It is supposed to block airflow so the cpu that is installed doesn't get bypassed. It has been about 15 years since I saw one of these, but I'm pretty sure they came with it in a single cpu configuration.
Oh yeah that makes sense, I was wondering about that. It'd be really unlikely they would've shipped that without a cover!
7063-cr2 ftw!!
20:17 -- Very Vice Grip Garage-y comment there 😅😅
oh yeah 😂
System p5? that seems huge than my HP Compaq Small Form Factor from 2008!
I do not miss my p5 days.
I find these IBM servers impressive, but according your experience you would tell that these are hard to work around?
I wouldn't say hard, but can be tricky to find the right documentation about the older ones. have to remember these are nearly 20 years old, and I haven't actually used them in any sort of production setting
One of the nics are gigabit, and my experience are that it's fast as heck. NFS to a modern storage solution are much faster than the internal SCSI-drives.
These were were often running Oracle databases and Websphere apps
ive never got a chance to play with a 1u server this was years ago im not sure how long ago exactly but i can remember it was a 1u Dell server in a silver color
deff got to fined a Quad core CPU for that Beast.
agreed!
cool video, thanks for sharing, there is documentation out there but first time I have seen a video about how HMC is actually installed. did you end up having to configure 'private network' in order to add the p5 system at the end?
I just plugged them right into my existing network and assigned them valid IPs
Back in the days, I've used smitty a lot.
Those PSUs have some Finnish (and Swedish) on them ("must be plugged to grounded outlets"), did it make its way to you all the way from here? Interesting :)
41:57 it needs another CPU installed in order to create this other LPAR. Once another is installed (matching the other CPU, or upgrade both of them to the max that model supports), I'd believe that would increment to 1 and then it would work. If it has only one socket, then this must be IBM lingo for cores. It looks like two and four core POWER5+ CPUs existed
Yeah this one is unfortunately a single-socket system with a two core CPU.
@@clabretro you also mentioned that half cores could be assigned, could existing LPAR be split up or something or free up something?
Exactly, that's what I'm thinking I'll try. I have some more RAM to put in there now and I'll try configuring the first AIX instance to not take up all the CPU cores.
Nice video. Refreshing my memory. I worked for 12 yrs from 2007 to 2019 on AIX n Redhat Linux. last 4 yrs I was part of some environment startup which failed. I am trying to get back to AIX/Linux job. But unable to crack interviews as I lost touch. I am US citizen of Indian origin. If any positions are available even a Junior position, please let me know.
Methinks clabretro enterprises needs a VAX 6000 or an AlphaServer GS160. :)
ha hopefully someday!
You need some intense power for a VAX 6000. Maybe three phase although I think that's just for certain tape drives.
careful. if you shrink that shrink tube, it's gonna cause a short. happened to a friend :P
yeah that's why I never shrank it haha. I need to make a better one, this one was a quick-and-dirty one just to test that MRV when I first got it (and I'm still using it, of course)
I think the X336 remote management is not working because the management card is actually missing. That card would be called Remote Supervisor adapter 2 slim. At least I couldn't see one of these installed. I also had to install one of these into my X3250. On that one I found out you can use the RSA II Slim from any other server, you only have to flash the correct firmware to work properly.
Huh, Logical Partitions sure do seem a lot like virtual machines...
Very similar. LPARs have some interesting stuff though, like partial CPU assignment.
I'm guessing you bought the machine from somewhere in Fennoscandia, since the PSU labels are in SWE/NOR/FIN :D nice machine!
That p5 came from a craigslist lot, the other gear was from Wyoming in the United States. but it's hard to say where it all really came from!
From the licensing screen at 42:12, there's some info that I remember off the top of my head:
- no Active Memory Sharing: this means no overcommitting memory, total allocated between active LPARs has to be under the physical maximum
- no IBM I Capable: there's a part of me that's sad by it, I'd love for you do a video on it!
- CoD Capable: that's from back in the day where they'd ship you more physical CPUs than you had ordered/licensed, and you could pay-per-time to use them (HMC would generate billing reports, you were contractually obligated to provide them to IBM) -- i'm half suprised a type 52A would have this on o.O
- no Micro-partitioning capable: IIRC this is one of the sad ones, this is the feature that lets you do fractional CPU allocation. I *think* without it you're limited to 2 LPARs (one per physical CPU you've got). You may still be able to create them, but it won't let them activate them.
- no Virtual I/O Server: this is an LPAR you can create, usually the first one you would (or two, on larger systems), that would own the physical resources (hard drives, ethernet cards), and then vend "virtual" resources to the other ones. It's not that common on a --52A, because it's a smaller machine, but this could've been a whole video in itself too! -- btw, that'd show up as"virtual resources" under Tasks > configuration
- Shared Ethernet Adapter Failover Capable: along with VIOs, this is a way to have a virtual ethernet bus between the VIO and the LPARs. A rough comparison would be SR-IOv today.
- Redundant Error Path Reporting Capable: Power machines are REALLY sturdy, and some components are not only redundant, but able to report on each other when the redundancy is activated. IDK if it'd be useful on a -52A.
- GX Plus Capable: that big multi-pin slot under the FSP, that's for it
- Active / Inactive partition mobility: with multiple (identical) machines under the same HMC, and a proper VIO setup, the HMC can orchestrate the migration of a host from one to the other -- even live!
That's what I can see from the scroll, if you want to write down any others that weren't shown, @ me and i'll see what I remember!
Thanks so much for this extra info, it's really useful. It's hard to find information about this online except from people like you that are already familiar!
I was also sad to see not IBM I capable, was looking forward to trying that. That's a bummer about no fractional CPU partitioning, that's one of the most fascinating parts about LPAR. But hopefully I am able to create at least two LPARs, that'd still be great.
How about getting hold of an old Timeplex Link 1 or GDC Meganux 1258 from the 80s.
9:00 The way I read that bulletin means that it's supported only in BOTH slots, not in only one of them. So, the official configuration requires two of those cards, and then the problem will not occur. Looks like someone pilfered one of your cards, if that's how you received the system.
Since these are shared 133MHz slots, I wouldn't be surprised if there's some weird timing issues and the clock is actually split between both slots.
Yeah came that way... suspicious thing is the other slot had its PCI blank cover on, usually those disappear when things are pilfered. Who knows though!
I wonder seeing this if there is a way to install mvs or z/os on this thing, since you have on the hmc the same behavior from the hmc I had to use to manage z/os
What type of RAM do you need for this? If it's typical server RAM, I'm assuming registered ECC. I have a little left over from a kit I bought for one of my servers that you are welcome to if you'd like.
It's 533mhz DDR2, if you're up for it you can email me at the email in the channel's about page!
Ouch! I understand that it was the Enterprise wilderness and money milking. But just today in a new installation probably no one would pay any bit of attention to think to get any of commercial Unix anymore. At least RHEL(clones) install themselves without much headache, of course management console is another paid ting either from RH or Canonical, but at least your start your server initialization much more easier.
yea fans are easily removable because dust.
14:31 the DC LED have a reversed icon. Is that just a different icon, or that means the polaris of output are also reversed between two PSUs ?
probably just a different icon, otherwise it'd be real unhappy
At what price were you buy ThinkCentre thin clients?
And also, what can you advise for a small (not about the size) home server? The budget is about $350
Look for the best CPU at the lowest wattage; These days home labs can be expensive to run power wise if you don't consider it.
@@TradieTrev I wanna more performance than low wattage
Enjoyed this! Am interested in using old servers as powerful home workstations.
Thanks! They're fun, but the trick with these older machines is they use a ton of electricity.
Could the hard drives that displayed 0MB capacity be from part of a RAID array?
hm I don't think so, this one doesn't have a RAID controller. could be related though.
the most cursed thing is you using a heat shrink but not actually shrinking it on that rj45 adapter
the wire job was so bad I never shrank it because it might mess things up 😆. been wanting to make new ones for a long time (these were just for testing) but they just keep on workin'
@@clabretro nothing more permanent than a temporary solution that's still working
Wonder if someone hit that zero mb with a Degauser since that would ruin the sector formating.
oh could be. whenever I get around to reviving them if they're permanently broken that could be the case
How loud and power hungry are they actually, when running 24/7 both?
The 336 was louder than the p5 (1U servers usually are), but I didn't measure DBs. They both pull well over 100 watts, I'll measure that stuff next time!
License to RAM upgrades? Is it IBM or Tesla? :o
IBM does not only have a setting for language but for "Cultural Convention" as well.
Where could I download the software IBM Hardware Management Console so I could create one in a virtual machine with Linux?
As far as I know IBM hosts them all via FTP here: public.dhe.ibm.com/software/server/hmc/recovery_images
EDIT: oh, and via HTTP, it would seem.
My I ask for hmc images?
First