climateprediction.net (CPDN) home page
Thread 'Running 32-bit MacOS Tasks on Linux with KVM'

Thread 'Running 32-bit MacOS Tasks on Linux with KVM'

Questions and Answers : Unix/Linux : Running 32-bit MacOS Tasks on Linux with KVM
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · Next

AuthorMessage
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65183 - Posted: 18 Feb 2022, 18:35:41 UTC - in response to Message 65182.  
Last modified: 18 Feb 2022, 18:41:18 UTC

Did you try the ignore_msrs incantation early in the instructions?

Another thing to be aware of, if you're using multiple machines: BOINC struggles to tell multiple VMs apart, if they're started at the same time with the same MAC address. Change the MAC address in the launch scripts, or just let one run for a few hours, get a full load of tasks, and then start another one. If they're identical MAC addresses, started around the same time, they get merged into each other and you end up with some "Abandoned" status tasks, as it seems a machine is requesting tasks, then saying it has no tasks and requesting more. I dropped a few on the floor that way (sorry!), though it's not hard to avoid once you figure out the root cause. Once they're split into separate machines, there are no further issues with that.
ID: 65183 · Report as offensive     Reply Quote
Profilegeophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2187
Credit: 64,822,615
RAC: 5,275
Message 65184 - Posted: 19 Feb 2022, 6:07:14 UTC - in response to Message 65172.  

@SolarSyonyk

You can't run the system device updates, they fail - so just select the point release, if you care. Or ignore it, it's a compute VM.

For some reason mine updated fine. Don't ask me why...

qemu/kvm lie to the guest as needed about the actual nature of the host CPU. As long as it supports the needed extensions (of the SSE and AVX variety), it should work totally fine. A new Ryzen ought to support everything needed to do this, and I expect will rock those units right along!

My Ryzen is identified by BOINC as a Penryn Core 2. In Kholia's notes link, he says you can enable AVX2 instructions and in the process have the CPU identified as "Haswell". But he writes about boot-clover.sh and I can't find that anywhere in the KVM directory tree. The link he gives about more information is out of commission and has been since 2019. I realize we don't need AVX2 to run these models, but if this can be done easily, I'd rather have the CPU identified in BOINC as Haswell rather than the ancient Penryn.

I'm running 2 hadcm3s now at 0.46 sec/TS on my Ryzen 7 3700X, along with 4 ARPs (usually) as I help mop up the resends in WCG before it shuts down on the 28th.
ID: 65184 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65185 - Posted: 19 Feb 2022, 7:39:24 UTC
Last modified: 19 Feb 2022, 7:46:39 UTC

Did you try the ignore_msrs incantation early in the instructions?
Going back to re-read,

Did carry out that step.

No other VM running.
ID: 65185 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65186 - Posted: 19 Feb 2022, 17:15:17 UTC - in response to Message 65185.  

OK, still same problem, will try again from the start.
ID: 65186 · Report as offensive     Reply Quote
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65187 - Posted: 19 Feb 2022, 17:21:38 UTC

What's your hardware, exactly? I've got it working on 5th, 6th, 8th gen Intel and an AMD box, but I haven't tried it on anything particularly weird.
ID: 65187 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65188 - Posted: 20 Feb 2022, 7:19:21 UTC - in response to Message 65187.  

What's your hardware, exactly? I've got it working on 5th, 6th, 8th gen Intel and an AMD box, but I haven't tried it on anything particularly weird.


CPU type AuthenticAMD
AMD Ryzen 7 3700X 8-Core Processor [Family 23 Model 113 Stepping 0]
Number of processors 16
Coprocessors NVIDIA NVIDIA GeForce GT 710 (978MB) driver: 470.99 OpenCL: 3.0
Virtualization Virtualbox (6.1.26_Ubuntur145957) installed, CPU has hardware virtualization support and it is enabled
Operating System Linux Ubuntu
Ubuntu 21.10 [5.13.0-28-generic|libc 2.34]
BOINC version 7.19.0
Memory 31.18 GB

I am still thinking there is something obvious that I am missing. My hardware shouldn't be a problem based on what I have read. And it is annoying me. I can build my own BOINC client and manager from source so this shouldn't be as challenging as it it is for me! I started again from the start with extracting the zip and then followed the readme but still getting the same problem.
ID: 65188 · Report as offensive     Reply Quote
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65189 - Posted: 20 Feb 2022, 15:10:26 UTC

The scripts shouldn't care about Virtualbox, they're using KVM.

About the only difference between your setup and my homeserver is that I'm running 20.04 LTS vs 21.10. I'm not sure that should make a real difference in virtualization, though.
ID: 65189 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65195 - Posted: 23 Feb 2022, 12:19:53 UTC

Still no joy here. I have tried the diagnostics script but that doesn't seem to show any clues at least to me.

~/OSX-KVM$ ./run-diagnostics.sh 
== Distro Info ==
Distributor ID:	Ubuntu
Description:	Ubuntu 21.10
Release:	21.10
Codename:	impish

== Loaded Modules ==
kvm_amd               126976  6
kvm                   876544  1 kvm_amd
ccp                   102400  1 kvm_amd

== Installed Binaries ==
QEMU emulator version 6.0.0 (Debian 1:6.0+dfsg-2expubuntu1.1)
Copyright (c) 2003-2021 Fabrice Bellard and the QEMU Project developers
virt-manager version 3.2.0
Python 2.7.18

== Networking ==
Found virbr0. Good.

I was wondering if there were any log files that could give a clue as to what is happening and if so where to find them?
ID: 65195 · Report as offensive     Reply Quote
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65196 - Posted: 23 Feb 2022, 14:35:19 UTC
Last modified: 23 Feb 2022, 14:36:20 UTC

The screen with the three icons with arrows under them - when you select the Base Image one and hit enter, it churns for a little bit and then reboots completely and ends up back there? Just trying to make sure I fully grasp the problem. Or is it just sitting there, doing nothing? If so, use the arrow keys and enter.

Are there any extra messages in the qemu window after it's done that?

Also, apparently CPDN is currently offline, power outage somewhere, and nothing but the website is up. I've got a few systems with half-downloaded tasks idling around waiting to finish downloading and finish uploading.
ID: 65196 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65197 - Posted: 23 Feb 2022, 15:30:43 UTC

The screen with the three icons with arrows under them - when you select the Base Image one and hit enter, it churns for a little bit and then reboots completely and ends up back there? Just trying to make sure I fully grasp the problem. Or is it just sitting there, doing nothing? If so, use the arrow keys and enter.


I select the base image one and hit enter. It flashes off, some text flashes on the screen too fast for me to read anything then it ends up back there. It shows as using about 106%cpu so just over one core but sits there for hours doing that. I don't see any changes in the OSX-kvm folder to give me a clue as to what is happening but given that even after leaving it for 5 hours on a Ryzen7 there is no change, I can only assume it is not installing. Hence my wondering about a log to look for clues.
ID: 65197 · Report as offensive     Reply Quote
Profilegeophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2187
Credit: 64,822,615
RAC: 5,275
Message 65198 - Posted: 23 Feb 2022, 16:00:36 UTC - in response to Message 65197.  
Last modified: 23 Feb 2022, 16:02:02 UTC

I select the base image one and hit enter. It flashes off, some text flashes on the screen too fast for me to read anything then it ends up back there. It shows as using about 106%cpu so just over one core but sits there for hours doing that. I don't see any changes in the OSX-kvm folder to give me a clue as to what is happening but given that even after leaving it for 5 hours on a Ryzen7 there is no change, I can only assume it is not installing. Hence my wondering about a log to look for clues.

Not sure if it will help or work, but perhaps you can capture the process where the text flashes on the screen using screen recording software. You can play it back using VLC or something and step through the frames to see if you can read the text? The only screen recording software I've used is Gtk-recordmydesktop but there are a bunch of probably better ones nowadays.
ID: 65198 · Report as offensive     Reply Quote
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65199 - Posted: 23 Feb 2022, 16:42:32 UTC

It should launch into the installer at that point.

I've done it successfully on 4 local systems, and two cloud boxes, using the steps provided - though only one was an AMD box. Not sure, sorry.
ID: 65199 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65200 - Posted: 23 Feb 2022, 16:52:27 UTC - in response to Message 65198.  

Thanks George, just installed ScreenRec and have started from scratch to see what happens.
ID: 65200 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65201 - Posted: 24 Feb 2022, 11:43:02 UTC

First screen recorder I tried, didn't like the vm. Second one gives

End SetConsoleMode
Start LoadKernelFromStream
End LoadKernelFromStream
Start InitRootStruct

root device uid is '24EA2003-EAD0-AE9D-EB255AB6BC95'
End InitBootStruct
Start LoadRamDisk
End LoadRamDisk
Start FinalizedBootStruct
Start RandomSeed
End RandomSeed


None of which gives me any clues whatsoever!
ID: 65201 · Report as offensive     Reply Quote
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65205 - Posted: 24 Feb 2022, 19:49:29 UTC

Indeed not. Sorry, I'm well and truly out of ideas.
ID: 65205 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65207 - Posted: 24 Feb 2022, 20:32:41 UTC - in response to Message 65205.  
Last modified: 24 Feb 2022, 20:42:07 UTC

no worries, looks like I might have some time to play with it before Linux work comes back on stream anyway.

Edit: Putting some of the output into a search engine suggests that the problem is that some AMD cpu's not being correctly identified or their features not being so leads to a kernel panic. I need to do more reading of the links I found and probably some more learning in general to understand it all before I have any idea if there is anything I can do about it.
ID: 65207 · Report as offensive     Reply Quote
SolarSyonyk

Send message
Joined: 7 Sep 16
Posts: 262
Credit: 34,915,412
RAC: 16,463
Message 65208 - Posted: 24 Feb 2022, 20:44:02 UTC

That's as good a guess as any others. My VMs are running out of WUs, so I'll have to find yet something else for them to do... (yes, I saw it was supposed to be fixed, no, uploads/downloads aren't working for me yet).
ID: 65208 · Report as offensive     Reply Quote
Profilegeophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2187
Credit: 64,822,615
RAC: 5,275
Message 65209 - Posted: 24 Feb 2022, 21:47:27 UTC - in response to Message 65207.  

no worries, looks like I might have some time to play with it before Linux work comes back on stream anyway.

Edit: Putting some of the output into a search engine suggests that the problem is that some AMD cpu's not being correctly identified or their features not being so leads to a kernel panic. I need to do more reading of the links I found and probably some more learning in general to understand it all before I have any idea if there is anything I can do about it.

Dave,

I'm running the same CPU you are. It's on a Gigabyte AX370m motherboard with 32 GB of RAM. Ubuntu and the installed VM are on a fast mechanical hard drive. Had a clean install of Ubuntu Mate 20.04 LTS on it in late January. I didn't do anything special to it although certainly the instructional steps are written by someone assuming we know something about VMs in Linux and about Macs. That's not to criticize. It's written pretty well, but for a first timer...
ID: 65209 · Report as offensive     Reply Quote
ProfileDave Jackson
Volunteer moderator

Send message
Joined: 15 May 09
Posts: 4541
Credit: 19,039,635
RAC: 18,944
Message 65211 - Posted: 25 Feb 2022, 11:05:32 UTC

certainly the instructional steps are written by someone assuming we know something about VMs in Linux and about Macs.


I have used VB before in Linux and at least know my way around the basics. I have never used a Mac however. The fact that it works for you means there is probably something that when tracked down will be really obvious but so far isn't to me.
ID: 65211 · Report as offensive     Reply Quote
Profilegeophi
Volunteer moderator

Send message
Joined: 7 Aug 04
Posts: 2187
Credit: 64,822,615
RAC: 5,275
Message 65217 - Posted: 27 Feb 2022, 19:38:54 UTC

Well, that was discouraging. We had a power cut this morning. Bringing up my linux PC, then the MacOS VM, 3 of the 6 tasks that were running (and 13 model months along) immediately crashed.
ID: 65217 · Report as offensive     Reply Quote
Previous · 1 · 2 · 3 · 4 · Next

Questions and Answers : Unix/Linux : Running 32-bit MacOS Tasks on Linux with KVM

©2024 cpdn.org