If you're going to do a fresh install, try using the GUI software manager to install BOINC. Richard seems to have had success with it.

Just to share info, BOINC 7.18.1 (client only, don't have the manager) works on Ubuntu 22.04 on WSL2.

I used the software manager on Mint but for some reason failed. However, I was successful on Zorin so far. So, I will stick to Zorin.
Now the 32-bit libs.

I believe Zorin is based on Ubuntu so you should only need one 32-bit package: lib32stdc++6. Run the following and try to get a task to test things:

sudo apt install lib32stdc++6

I believe Zorin is based on Ubuntu so you should only need one 32-bit package: lib32stdc++6. Run the following and try to get a task to test things:

sudo apt install lib32stdc++6

It is based on Ubuntu. After all this exercise, I do not think GPUGrid, multi-core plus GPU and CPDN with OpenIFS can get along together. @Richard, what conclusions?

I do not think GPUGrid, multi-core plus GPU and CPDN with OpenIFS can get along together. @Richard, what conclusions?

Why do you say that, what exactly do you mean by 'not get along'? Each of those projects will require resources on your machine. If you can't run them concurrently, then adjust your boinc client accordingly and only run a few at a time. There is no technical reason why independent projects can't work on the same machine unless the computer doesn't have the necessary resource (typically RAM). I'm interested to know why you think they don't.

I run GPUGrid, and that particular MT+GPU app has its quirks. But with understanding and management - I've posted a few tips - they chug along quite happily.

I haven't seen a real-life IFS task yet, but when they arrive I'll run a few while keeping the rest of the machine relatively lightly loaded. I imagine I'll reach a similar conclusion.

I won't try running both apps together until I've totted up the needs of both (cores, RAM) and compared with what I've got. My machines are babies in this context (maximum 6 cores), but I reckon I could run one of each task type at the same time, plus one or two of my regular lightweight gap-fillers (Einstein, NumberFields).The trick will be to teach BOINC to allocate the right resources for each job, and then sit back to watch the fun.

I used the software manager on Mint but for some reason failed.

You need that reason - you have to dig, or no-one can help you. If things are getting tough, keep away from the cutting edge - Mint 20.3 is stable and mature, so try that.

Reboot the system and then try again.

I tried plus tried to reinstall Boinc Manager and Boinc Client. It reported back, you already have the latest version but I cannot see the GUI. I am going to do a fresh install of Mint and then try again.

A fresh install of Mint is probably not going to help you, just waste your time. Can you be a bit more specific and provide the responses from your terminal session so we can better help.

When you say 'cannot see the GUI' what exactly do you mean? Did you install just the client or the boinc-manager as well?
If it says you have the latest version, locate them. Try the command:

which boinc

you should get something like:

/usr/bin/boinc

The boincmgr (GUI) app will be in the same directory (yours may differ, use the directory from the output of above command):

ls -l /usr/bin/boincmgr

If it's not there I suggest you have not actually installed the GUI manager at all. Try to install it again.

I think we are drifting off topic for this particular thread. If you are still having problems, I suggest starting a new thread on the message board and we can continue there.

Cheers.

you should get something like:

/usr/bin/boinc

I get,

which boinc
/usr/local/bin/boinc

but, I downloaded the source of the latest master from git-hub and compiled it myself. Not sure if it is those in charge of package management who choose where to put it if you install the packaged version?

I do not think GPUGrid, multi-core plus GPU and CPDN with OpenIFS can get along together. @Richard, what conclusions?

Why do you say that, what exactly do you mean by 'not get along'? Each of those projects will require resources on your machine. If you can't run them concurrently, then adjust your boinc client accordingly and only run a few at a time. There is no technical reason why independent projects can't work on the same machine unless the computer doesn't have the necessary resource (typically RAM). I'm interested to know why you think they don't.

I have babies as Richard said. Six core machines with sixteen GB RAM. Each GPUGrid task is taking up a maximum of 98% CPU at peak plus total, machine and task 11.2 GB of RAM. I do not know what OpenIFS will do.
I am quite comfortable with Zorin, got it set up but VirtualBox with Zorin and a task from Einstein peaked the machine out. One task of Moo Wrapper can run comfortably with a Python in the cage without VB. Different projects, different requirements. This multi-tasking is giving a pretty high error rate at GPUGrid. My experience, not quoting.
Richard, computers came along pretty late for me, somewhere in 1980. Before IBM PC, my experience was with Texas Instruments 8086, and 8088. GWBasic was good enough for loading calculations. I just do not have the know-how to play with these but I can follow instructions if someone has the time to teach.
How do I adjust Boinc Client? Any instruction or Cheat Sheet somewhere in Boinc'dom?

This multi-tasking is giving a pretty high error rate at GPUGrid.

I have never found error rates go up as a result of overtaxing my computers. They do sometimes become unresponsive to mouse and or keyboard. I have noticed massive slow downs when I don't have enough RAM to run without a lot of recourse to swap.

How do I adjust Boinc Client? Any instruction or Cheat Sheet somewhere in Boinc'dom?

Not that I am aware of. There are the Man pages for BOINC but they tell you about each individual change you can make rather than about how they play together. Those are things that by and large are try it and see what happens or trawl through lots of often contradictory posts on the forums. The trouble is that there are so many different projects and settings that when you start combining them the number of possibilities rockets. Before I first attempted to build BOINC from source I thought it would be a matter of just following instructions. As it was, I found myself doing down a rabbit warren of unmet dependencies and having to work out what package contained them.

Even now OS upgrades often mean I have to change something to get the latest version to compile. The best you can do is research what you want to do on the forums both the BOINC ones and the projects you run. If you find your experience doesn't match that of others say so and there is a better than even chance someone might have a clue as to why!

Even now OS upgrades often mean I have to change something to get the latest version to compile. The best you can do is research what you want to do on the forums both the BOINC ones and the projects you run. If you find your experience doesn't match that of others say so and there is a better than even chance someone might have a clue as to why!

And therein lies the problem with boinc. It was meant to be an easy-to-use application to load onto volunteer machines which people could forget about -- it just ran as a nice screensaver. Now, unfortunately it can often require a degree of technical skill (and interest) to install & manage.

Six core machines with sixteen GB RAM. I do not know what OpenIFS will do.

OpenIFS at the resolutions we run now will work with 16Gb RAM. They have similar resource demand to a LHC Atlas job.

You won't be able to run higher resolution OpenIFS but remember that your machine wouldn't get them anyway. The tasks will have minimum requirements. If your machine doesn't satisfy those, it won't get the tasks. That's why I wasn't sure what you meant about 'not playing nicely'. The project servers should not be sending you tasks your machines can't handle.

Each GPUGrid task is taking up a maximum of 98% CPU at peak plus total, machine and task 11.2 GB of RAM.

If those are too high you can change the computation settings in boincmgr and then you won't get tasks with high RAM demands.

My experience, computers came along pretty late for me, somewhere in 1980.

I'd say that was quite early! And not long after me!

How do I adjust Boinc Client? Any instruction or Cheat Sheet somewhere in Boinc'dom?

I would strongly advise you not to twiddle the settings too much. The forums are full of posts by people who have changed settings and suddenly things aren't working anymore. Follow the old adage 'if it ain't broke don't fix it' and only change something if you spot a problem.

And therein lies the problem with boinc. It was meant to be an easy-to-use application to load onto volunteer machines which people could forget about -- it just ran as a nice screensaver. Now, unfortunately it can often require a degree of technical skill (and interest) to install & manage.

I'd like to ever-so-gently push back on that one. I think, in general, that BOINC can still be useful straight out of the box, with no actual "need" to twiddle anything. Though I do have a suspicion that some people 'download' BOINC, but don't realise that you then have to 'install' it before it does anything useful - as they do with anti-virus programs. I had that added to the download page deliberately...

Bear in mind that there is a huge range of variation in the human volunteers who run BOINC, and in the computers they bring to the party. The people you meet here are a tiny, tiny, minority of the BOINC population at large. We are the curious ones, we are the inquisitive ones, we (some of us) are the competitive ones. We like to lift the hood, and we like - no, we love - bright shiny new things. And we play with things until they break.

I think it's more of a problem for the bigger projects, who would like to leverage the enhanced features of the newest hardware: I think CPDN is in that group. Another is GPUGrid, where we're struggling with a newish Windows app which can require that you fiddle manually with the Windows Virtually Memory (paging file) settings. Those are buried three levels down in the advanced section of the system control panel. That's a big ask, but it's a project ask, not a BOINC ask.

Like others here, I bought my first micro computer around 1980 or 1981 (I still have it!). But I was exposed to mainframe computers, literally since before I was born in 1952. My parents both worked with the original EDSAC machine at Cambridge University: my father may even have helped build it out of war surplus parts, and my mother wrote of pulling all-night shifts alone in the old Maths Lab, so she could get some serious work done on her PhD project while I was at home with the au pair. Later, I spent a full year in the lab's successor, being taught "computer science" - the full gamut, not just programming. But after that, I left the field completely alone, until the scene changed from computers being things "they use to get our gas bill wrong", to being things we can use to help put the world to rights. I think that gap helps me to keep a foot in both camps - the community out there, and the technical club in here.

My history is of a similar vintage. '73/4 I was learning algol on Eliott22000 mainframe. A few years later I went in a completely different direction and was till taking early retirement, in mental health nursing for many years though during that time, I bought my first Tandy PC with 128K RAM. Still a step up from a ZX80 that I had typed programs into with BASIC or sometimes printed out machine code from magazines! Next PC was a 286 that I built myself and then I got into running BOINC for CPDN sometime near the start with a different username and email that got lost somewhere along the way. I defenestrated in 2000 when if you didn't like playing with shiny or not so shiny new things Linux was shall we say, not the easiest thing to install!

I don't do it so much now but I did spend several years following the fix it till it breaks then fix it philosophy. About once every other week or so, I upgrade BOINC to the latest master. Most times I never notice what has changed.

Each GPUGrid task is taking up a maximum of 98% CPU at peak plus total, machine and task 11.2 GB of RAM.

I am not clear what you mean. I run a bunch of Boinc tasks at a time (currently 12), and each processor (core) takes 99% of the cpu time of the cpu it is running on. This is how it should be, and no big deal. Even so, there are three other processors that have almost nothing to do.

At the moment, each einstein process takes 1.2% of my RAM, each CPDN task takes 1% of my RAM, and the WCG tasks take far less. Rosetta does not condescend to send me any work units lately. My RAM is 64 GBytes.

top - 16:46:34 up 5 days,  4:01,  1 user,  load average: 12.51, 12.55, 12.48
Tasks: 469 total,  13 running, 456 sleeping,   0 stopped,   0 zombie
%Cpu(s):  0.3 us,  0.1 sy, 74.4 ni, 24.9 id,  0.0 wa,  0.2 hi,  0.1 si,  0.0 st
MiB Mem :  63772.8 total,    986.2 free,   9148.3 used,  53638.4 buff/cache
MiB Swap:  15992.0 total,  15794.0 free,    198.0 used.  53828.4 avail Mem 

    PID    PPID USER      PR  NI S    RES  %MEM  %CPU  P     TIME+ COMMAND                                                                  
 466564   16165 boinc     39  19 R 765752   1.2  99.4  7 201:55.60 ../../projects/einstein.phys.uwm.edu/hsgamma_FGRP5_1.08_x86_64-pc-linux+ 
 466098   16165 boinc     39  19 R 765504   1.2  99.4  0 211:33.92 ../../projects/einstein.phys.uwm.edu/hsgamma_FGRP5_1.08_x86_64-pc-linux+ 
 456719   16165 boinc     39  19 R 765500   1.2  99.2  1 362:49.34 ../../projects/einstein.phys.uwm.edu/hsgamma_FGRP5_1.08_x86_64-pc-linux+ 
 409041  409034 boinc     39  19 R 674544   1.0  99.4  3   1225:58 /var/lib/boinc/projects/climateprediction.net/hadsm4_um_8.02_i686-pc-li+ 
 431436  431428 boinc     39  19 R 670332   1.0  98.9 13 831:48.37 /var/lib/boinc/projects/climateprediction.net/hadsm4_um_8.02_i686-pc-li+ 
 442269  442262 boinc     39  19 R 670304   1.0  99.4 15 611:45.87 /var/lib/boinc/projects/climateprediction.net/hadsm4_um_8.02_i686-pc-li+ 
 442325  442318 boinc     39  19 R 668472   1.0  99.1  4 610:21.26 /var/lib/boinc/projects/climateprediction.net/hadsm4_um_8.02_i686-pc-li+ 
 444454  444449 boinc     39  19 R 668012   1.0  98.9  0 565:47.04 /var/lib/boinc/projects/climateprediction.net/hadsm4_um_8.02_i686-pc-li+ 
 471817   16165 boinc     39  19 R 190472   0.3  99.4  2 124:20.29 ../../projects/www.worldcommunitygrid.org/wcgrid_opn1_autodock_7.21_x86+ 
 474668   16165 boinc     39  19 R 156116   0.2  98.9  6  93:51.00 ../../projects/www.worldcommunitygrid.org/wcgrid_opn1_autodock_7.21_x86+ 
 478188   16165 boinc     39  19 R 127932   0.2  99.2  5  41:00.76 ../../projects/www.worldcommunitygrid.org/wcgrid_opn1_autodock_7.21_x86+ 
 476894   16165 boinc     39  19 R  73000   0.1  99.7 11  52:24.55 ../../projects/www.worldcommunitygrid.org/wcgrid_mcm1_map_7.61_x86_64-p+
  16165       1 boinc     30  10 S  43900   0.1   0.1 13  76758:02 /usr/bin/boinc     <---<<< This is the boinc client

---

GPUGrid is slightly different(like CPDN will be with OpenIFS). It is not a single-thread project or even a single-core project. GPUGrid takes over the complete processor. If you have a 64-core processor, it will take up all of them. It graciously leaves one core alone for the OS. As to OpenIFS, I cannot find much as to how it will operate. Will it be multi-threaded, which I have read somewhere it will be? If it is multi-threaded and so is GPUGrid, that will be quite a party going on unless I build a fence around them. Now it starts going beyond my capability and I do not think a six-core machine will be sufficient.
https://github.com/CPDN-git/openifs
https://confluence.ecmwf.int/display/OIFS/OpenIFS+User+Guide
https://confluence.ecmwf.int/display/OIFS/5.4+Download+OpenIFS
If you can point out a website where I can find out more, most welcome.
Another point was raised somewhere. If GPUGrid is taking over the complete processor and if I run another WU from some other project, the hypothetical question is will there not be interference?

Let's straighten this out.

1: OpenIFS is single core at the moment, will go multiple core in the future but not beyond 4 core.

2: you tell boinc what fraction, that is, how many cores & ram to use, and it will not give you tasks that go beyond that.

3: you have even finer control by using an app_config.xml file which allows you to restrict multicore apps to use even fewer cores than you make available to boinc.

I've not used GPUgrid but it sounds like a badly behaved application.

Hope that helps.

I've not used GPUgrid but it sounds like a badly behaved application.

It's a deliberate design choice, but it doesn't fit into the BOINC paradigm space very well.

It's an artificial intelligence application, using the pytorch machine learning python framework. It alternates between training and running AI agents (I think training on the CPU, then running on the GPU - but possibly vv). The Python package spawns huge numbers of threads, each allocating a lot of virtual memory - that's the bit Windows can't cope with - but each thread has pretty low utilisation. I find it runs better under Linux if I allocate 3 cores and 1 GPU via app_config: other projects can run alongside with no problems.

I have got everything ready but where do I find Boinc folders on Virtual Box? The config files that are being spoken about I could find them in FAT but I am lost on NTFS. On both, VB and Windows. Seems like some spoon-feeding is required. Boinc Wiki is silent. Where are the Project Directories?
PS. Please don't get annoyed, I am trying. Some background. I am a retired sailor with part-time quarry work so not much hands-on experience with computers.

Boinc usually puts its files in /var/lib/boinc-client but you won't see any project directory files until you join a project. When you do, there will be files in the subdirectory "project".

ps. Not annoyed. My father was in the royal navy. I do find forums a frustrating slow way to help people though