BOINC Pentathlon 2020

[Windows7ge]

7 minutes ago, leadeater said:

I did but it was nothing even close to what the others were dropping. Plus to do more than what I did I'd have to sacrifice on other projects which we needed. I just don't think we had the numbers to compete lol.

 

Need to get all those FAH folders over for the event, for how well we do on that we shouldn't do any less than 1st for GPU projects lol.

Unlike most people who contribute I don't have massive amounts of GPU power. Figuring out AMD drivers on Linux is kind of a PITA as well but it does result in greater numbers than on Windows.

 

I'll probably do as I did last year and mostly assist the Marathon. If Ithanul wants me to put a couple rigs or VMs on one project or another that's fine. I want to see us pull off Top 10 overall  again.

[Windows7ge]

I've been thinking of making a BOINC tutorial that covers how to setup the software across Windows and the most popular distributions of GNU/Linux (Debian(Ubuntu/Ubuntu Server/Mint/PopOS)/Arch(Manjaro)/RHEL&RHL(CentOS/Fedora). A quick forum search I see @Ithanul made a folding/boinc guide back in late 2017. Maybe it's a good time for an updated version?

 

Unfortunately I don't have access to an NVIDIA GPU and I've not even begun to figure out if AMD drivers exist for Arch/RHEL/RHL so besides Windows I think the guide would be CPU only. That or the GPU sections would have to be added later if others are willing to assist me.

[Rusted]

Got boinc docker running in my unraid server, it has only Xeon E3-1220v5 and can't use all cores for boinc. Need to see if I can get my test/junk machine running, I think that has i5-3450 so it should have more power than my server and it can use all cores to run boinc

[Windows7ge]

I was sooo close, gah this kind of ticks me off.

[leadeater]

7 minutes ago, Windows7ge said:

I was sooo close, gah this kind of ticks me off.

To 1m WCG points in a day?

 

Edit:

Also you seem to be doing a lot of smaller low run time jobs or something, you preferencing a specific project?

 

04/22/2020

1:075:11:28:56

1,686,254

3,090

04/19/2020

0:333:21:53:03

1,384,204

2,923

 

Sample of ones where my results returned is similar in number, looks like run time is the big points factor.

[Windows7ge]

9 minutes ago, leadeater said:

To 1m WCG points in a day?

 

Edit:

Also you seem to be doing a lot of smaller low run time jobs or something, you preferencing a specific project?

 

04/22/2020

1:075:11:28:56

1,686,254

3,090

04/19/2020

0:333:21:53:03

1,384,204

2,923

 

Sample of ones where my results returned is similar in number, looks like run time is the big points factor.

Yep.

 

I'm looking to get my Smash Childhood Cancer badge up higher:

 

 

It's the hammer. 5 year diamond right now. The jobs don't take as long as say Mapping Cancer Markers or Microbiome Immunity Project. I've noticed running SCC alone increased my daily output on BOINCstats as well.

[Ithanul]

On 4/23/2020 at 12:03 PM, Windows7ge said:

I've been thinking of making a BOINC tutorial that covers how to setup the software across Windows and the most popular distributions of GNU/Linux (Debian(Ubuntu/Ubuntu Server/Mint/PopOS)/Arch(Manjaro)/RHEL&RHL(CentOS/Fedora). A quick forum search I see @Ithanul made a folding/boinc guide back in late 2017. Maybe it's a good time for an updated version?

 

Unfortunately I don't have access to an NVIDIA GPU and I've not even begun to figure out if AMD drivers exist for Arch/RHEL/RHL so besides Windows I think the guide would be CPU only. That or the GPU sections would have to be added later if others are willing to assist me.

I do need to get around and re-look over that guide and do some updated screenshots.

 

I do have Mint Mate 19.2 loaded on my 2nd rig and have F@H installed on it.  But not BOINC, so I will see if my upcoming days off, I can get at least the BOINC part updated.

I do have Mint Mate 19.3 ISO downloaded and the newer CentOS, so may get around and do some fresh installs.

 

Maybe with some folks interested in doing setups in VMs, Docker, headless, etc.  can get all the info gathered up and consolidated into one guide.

 

For Windows, BOINC is pretty straight forward to install.  At most, the more tricky bits is like projects where one can fine tune the app config for the project to get better output.

[leadeater]

27 minutes ago, Ithanul said:

Maybe with some folks interested in doing setups in VMs, Docker, headless, etc.  can get all the info gathered up and consolidated into one guide.

I might switch over one of my VMs that's running docker to native and see if there is a performance difference. I've got two identical so I can run them side by side and limit to single WCG project. With how easy it is to spin up the Docker containers I don't think VMs are worth it anymore.

 

I've also recently added a guide for compiling BOINC and other necessary stuff for getting LHC native working so if you think it's necessary you can link to that, it's quite a process though so it'll put more people off just seeing it so yea "buyer beware" lol.

 

42 minutes ago, Ithanul said:

At most, the more tricky bits is like projects where one can fine tune the app config for the project to get better output.

Is there really much we can actually do with that to get higher output?

[Windows7ge]

48 minutes ago, Ithanul said:

I do need to get around and re-look over that guide and do some updated screenshots.

 

I do have Mint Mate 19.2 loaded on my 2nd rig and have F@H installed on it.  But not BOINC, so I will see if my upcoming days off, I can get at least the BOINC part updated.

I do have Mint Mate 19.3 ISO downloaded and the newer CentOS, so may get around and do some fresh installs.

 

Maybe with some folks interested in doing setups in VMs, Docker, headless, etc.  can get all the info gathered up and consolidated into one guide.

 

For Windows, BOINC is pretty straight forward to install.  At most, the more tricky bits is like projects where one can fine tune the app config for the project to get better output.

VMs would be my plan for doing all the testing and taking screenshots. I intend to do both head/headless & remote access options across the platforms.

 

6 minutes ago, leadeater said:

I might switch over one of my VMs that's running docker to native and see if there is a performance difference.

What hypervisor?

[leadeater]

15 minutes ago, Windows7ge said:

What hypervisor?

ESXi, that's all I use. I went with docker only because I want to try queuing up tasks then suspend the container once done then run up another container instance etc. Really is no different to a VM other than there is no OS management required for them.

 

I'm also maybe not doing the most optimal VM setup for the hardware, maybe. The VM spans across 3 sockets, I have all that stuff configured in the VM settings so it presents the NUMA stuff but 3 VMs sized and pinned to a socket might give better performance. If I were able to run these long term I'd go through the effort to find out.

 

15 minutes ago, Windows7ge said:

I intend to do both head/headless & remote access options across the platforms.

I just do everything via script/cli for multiple systems, it's just easier. Once a pent project is announced I create a script that tells each host to pause all current work then add/start that project then I just comment out each host/instance I don't want at the time of running it. So there's 5 scripts.

 

If you are going to be managing more than 2 or 3 BOINC instances I highly recommend the BOINC cli.

[Windows7ge]

24 minutes ago, leadeater said:

ESXi, that's all I use. I went with docker only because I want to try queuing up tasks then suspend the container once done then run up another container instance etc. Really is no different to a VM other than there is no OS management required for them.

 

I'm also maybe not doing the most optimal VM setup for the hardware, maybe. The VM spans across 3 sockets, I have all that stuff configured in the VM settings so it presents the NUMA stuff but 3 VMs sized and pinned to a socket might give better performance. If I were able to run these long term I'd go through the effort to find out.

 

I just do everything via script/cli for multiple systems, it's just easier. Once a pent project is announced I create a script that tells each host to pause all current work then add/start that project then I just comment out each host/instance I don't want at the time of running it. So there's 5 scripts.

 

If you are going to be managing more than 2 or 3 BOINC instances I highly recommend the BOINC cli.

I'm willing to test this using QEMU/KVM after the event is over. I'll setup hugepages for two VMs but I'll only pin CPU cores for one. The other will use whichever cores are free regardless of NUMA node then I'll just let it whirl till I see a discrepancy or if after say a month there's none.

 

I have...9~10 headless instances to manage. I enabled remote boinc-manager access on all of them but I did research how boinccmd works while testing out cloud servers. I have thought about how I could write a script to SSH into each server and just run a project update or change a setting. I will need individual VM control though as not all VMs are going to fill their queues at the same time. This makes remote boinc-manager useful.

[leadeater]

5 minutes ago, Windows7ge said:

I have thought about how I could write a script to SSH into each server and just run a project update or change a setting

You don't need SSH at all, BOINC has remote management and CLI all built in.

https://boinc.berkeley.edu/wiki/Boinccmd_tool

 

boinccmd [--host hostname] [--passwd passwd] command

 

[Windows7ge]

3 minutes ago, leadeater said:

You don't need SSH at all, BOINC has remote management and CLI all built in.

https://boinc.berkeley.edu/wiki/Boinccmd_tool

 

boinccmd [--host hostname] [--passwd passwd] command

 

Do you know if this works after editing remote_hosts.cfg or does it require another?

[leadeater]

4 minutes ago, Windows7ge said:

Do you know if this works after editing remote_hosts.cfg or does it require another?

Every BOINC client needs to allow which ever host you are running boinccmd on access, it comes through the same way as using the GUI boinc manager.

[Windows7ge]

1 minute ago, leadeater said:

Every BOINC client needs to allow which ever host you are running boinccmd on access, it comes through the same way as using the GUI boinc manager.

Alright, that should be all I need then. Any idea if there's a command to check the queue? How many jobs are still not completed?

[leadeater]

5 minutes ago, Windows7ge said:

Alright, that should be all I need then. Any idea if there's a command to check the queue? How many jobs are still not completed?

Yes but you might need to grep the output to filter out running tasks.

 

boinccmd --get_tasks

boinccmd --get_simple_gui_info

 

Run those and see, default is localhost so don't need to specify --host or --passwd

[leadeater]

Quote

Like many organisations, CERN is also affected by the COVID-19 pandemic. Researchers in the CERN community are trying to help out in different ways, as explained on this web page.

 

As part of this effort to fight COVID-19, we also contribute computing power to Folding@home and Rosetta@home from temporarily available servers that were about to be decommissioned.

 

During periods like this with little work from LHC@home, we also encourage you to participate in other BOINC projects such as Rosetta@home and contribute to the global fight of the pandemic.

 

Many thanks for your contributions to LHC@home and continued happy crunching!

 

With the best of wishes of good health for your and your families from the LHC@home team.

https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=5407#42258

 

So maybe no LHC this year then????

[Windows7ge]

@leadeater We're off to a good start:

 

 

Spoiler

Oh no I didn't hide muh IP's. 

 

[Ithanul]

4 hours ago, leadeater said:

I might switch over one of my VMs that's running docker to native and see if there is a performance difference. I've got two identical so I can run them side by side and limit to single WCG project. With how easy it is to spin up the Docker containers I don't think VMs are worth it anymore.

 

I've also recently added a guide for compiling BOINC and other necessary stuff for getting LHC native working so if you think it's necessary you can link to that, it's quite a process though so it'll put more people off just seeing it so yea "buyer beware" lol.

 

Is there really much we can actually do with that to get higher output?

I need to get around to learning about docker and doing its setup.  Where I work atm don't use such or even the equipment in the Guard unit I'm part of, but won't hurt for me to get some knowledge on that in case some day either looks into utilizing it.  Was funny enough when the main VMWare controller was switch to SUSE and most folks in my Guard unit hardly a clue how to fix it when during an exercise its filesystem went belly up.

 

Hmmm, ya got a link to that guide?  I be interested in reading through that.

 

[Windows7ge]

This weekend I'm going to work on getting PROXMOX setup on a spare SSD. The 5960X rig is too good to leave to only fill one queue. I'd do the same to the 3930K rig but I'm fresh out of SSDs at that point. The PSU it has is under powered so I don't know if I want to give it a beat-up spinning rust disk or not.

[leadeater]

2 hours ago, Ithanul said:

Hmmm, ya got a link to that guide?  I be interested in reading through that.

LHC native or the Docker? only wrote one for LHC native.

https://linustechtips.com/main/topic/1181316-lhchome-native-linux-guide/

 

 

Docker page is here though, you'll see what I mean by how easy it is once you try it. You do need to actually install Docker first.

https://hub.docker.com/r/boinc/client

 

[GOTSpectrum]

We are now up to 23 teams for the Pentathlon!

[Windows7ge]

Even though nobody asked I sucked it up and wasted my entire Saturday getting both the 5960X rig & 3930K rigs setup with Containers running on PROXMOX. I will be multi-instance bunkering on ALL fronts.

 

The 5960X rig was a PITA. Wouldn't recognize the GPU, then when it'd recognize another it'd be displaying 1920x1080 on a 800x600 output so it was unusable, then it didn't add the new drive to the boot menu so trying to go headless meant I had to put a head on it again just to fix that dumb BIOS setting. Miraculously the 3930K rig was quite cooperative besides the first disk I tried to attach kept reporting COMRESET errors and the PROXMOX installer couldn't see the drive. So that's dead. I dug up a 500GB WD Blue and it's working so far. No errors.

 

Gotta love old hardware. Sometimes I wonder if it's even worth it. Additionally I've lowered the queue on all my current rigs to 0.1 days. This way when the call to bunker comes I can finish my jobs and get bunkering faster.

[Ithanul]

I should try to get a few more of my hardware put together.  Though, my issue is only one air cooler my X99 platforms.  I got plenty of water blocks, but that means I have to bend tubes and bubble chase.

 

Though, eyes a 1950X over in a sales area on a forum.  That price is tempting me.  Also saw a 2920X being sold as well.  Of course, both being sold with mobos.  Wonder if either of those chaps be willing to sell just the CPU.  Unless I can find a good deal on an used 2950X as that what I want to drop in the main PC.

 

:I  Also.....darn GPU upgrade itch is hitting me now.  Need to get a few more sold off though before getting a new one.

[Windows7ge]

@leadeater

Use boinccmd he said,

it's more efficient he said.

 

Meanwhile amateur scripter here:

 

Spoiler

echo -n "win-hyper-v-bn1 queue.....: " && boinccmd --host 192.168.0.201 --get_tasks | grep -c no
echo -n "win-hyper-v-bn2 queue.....: " && boinccmd --host 192.168.0.202 --get_tasks | grep -c no
echo -n "win-hyper-v-bn3 queue.....: " && boinccmd --host 192.168.0.203 --get_tasks | grep -c no
echo -n "win-hyper-v-bn4 queue.....: " && boinccmd --host 192.168.0.204 --get_tasks | grep -c no
echo -n "proxmox-qemu-kvm-bn5 queue: " && boinccmd --host 192.168.0.205 --get_tasks | grep -c no
echo -n "proxmox-qemu-kvm-bn6 queue: " && boinccmd --host 192.168.0.206 --get_tasks | grep -c no
echo -n "ubuntu-qemu-kvm-bn7 queue.: " && boinccmd --host 192.168.0.207 --get_tasks | grep -c no
echo -n "ubuntu-qemu-kvm-bn8 queue.: " && boinccmd --host 192.168.0.208 --get_tasks | grep -c no
echo -n "5960x-bn9 queue...........: " && boinccmd --host 192.168.0.209 --get_tasks | grep -c no
echo -n "3930k-bn10 queue..........: " && boinccmd --host 192.168.0.210 --get_tasks | grep -c no

 

 

Spoiler

#!/bin/bash
echo -n 'win-hyper-v-bn1: '
if boinccmd --host 192.168.0.201 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
	echo 'Project Set'
else
	echo 'Project Not Set - Setting Project'
	boinccmd --host 192.168.0.201 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'win-hyper-v-bn2: '
if boinccmd --host 192.168.0.202 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.202 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'win-hyper-v-bn3: '
if boinccmd --host 192.168.0.203 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.203 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'win-hyper-v-bn4: '
if boinccmd --host 192.168.0.204 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.204 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'proxmox-qemu-kvm-bn5: '
if boinccmd --host 192.168.0.205 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.205 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'proxmox-qemu-kvm-bn6: '
if boinccmd --host 192.168.0.206 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.206 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'ubuntu-qemu-kvm-bn7: '
if boinccmd --host 192.168.0.207 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.207 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n 'ubuntu-qemu-kvm-bn8: '
if boinccmd --host 192.168.0.208 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.208 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n '5960x-bn9: '
if boinccmd --host 192.168.0.209 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.209 --project_attach http://www.worldcommunitygrid.org/ 
fi
echo -n '3930K-bn10: '
if boinccmd --host 192.168.0.210 --get_project_urls | grep -q "http://www.worldcommunitygrid.org/"; then
        echo 'Project Set'
else
        echo 'Project Not Set - Setting Project'
        boinccmd --host 192.168.0.210 --project_attach http://www.worldcommunitygrid.org/ 
fi

 

 

The first one will give me a print-out of how many jobs each VM has left to complete. The second will selectively add the World Community Grid project to the VMs that don't already have them. As I've found out the software isn't smart enough to tell you you already have the project added. It'll add it twice and that's problematic so you have to write the script to check if the project already exists.

 

boinccmd still appears to be limited in that it doesn't support changing settings such as Days of Work. I'll still have to write a script that goes into each VM and edits the global prefs override file from there I can use boinccmd to run the override command.

 

Or I can just use boinc-manager. At this point I think that would be less tedious.