I am running only 2 projects: Rosetta at 80% and Einstein at 20%.

For Rosetta, BOINC manager keeps "Requesting new tasks for Intel GPU," which ALWAYS results in "Scheduler request completed: got 0 new tasks."

Then Einstein comes along a few minutes later and scores new tasks by "Requesting new tasks for CPU."

This results in my queue filling up with Einstein tasks, even though I'm running Rosetta at 80%.

Is there a way to fix this?

---

Yes, you can setup a control file, that allows you to establish your preferences on a per project basis. I recently deciphered the instructions and tried to write instructions an actual person could follow. Let me know if I need to clarify anything in this post about it.

---

Rosetta Moderator: Mod.Sense

Thanks. Just for mental serenity, is this what I should see in the new Event Log?

6/7/2017 10:26:39 AM | rosetta@home | Config: excluded GPU. Type: intel_gpu. App: all. Device: all

I am running only 2 projects: Rosetta at 80% and Einstein at 20%.

For Rosetta, BOINC manager keeps "Requesting new tasks for Intel GPU," which ALWAYS results in "Scheduler request completed: got 0 new tasks."

Then Einstein comes along a few minutes later and scores new tasks by "Requesting new tasks for CPU."

This results in my queue filling up with Einstein tasks, even though I'm running Rosetta at 80%.

Is there a way to fix this?

Rosetta will only run on CPU.
Einstein will run on CPU or GPU AND ... the GPU version requires 1 full CPU.

Running Einstein on ONLY the GPU plus the GPU required 1 CPU ... will likely be more than 20%.

I would suggest you change the Einstein preferences to only run GPU jobs on your machine.

The idle CPUs will then request Rosetta jobs.

I am running only 2 projects: Rosetta at 80% and Einstein at 20%.

For Rosetta, BOINC manager keeps "Requesting new tasks for Intel GPU," which ALWAYS results in "Scheduler request completed: got 0 new tasks."

Then Einstein comes along a few minutes later and scores new tasks by "Requesting new tasks for CPU."

This results in my queue filling up with Einstein tasks, even though I'm running Rosetta at 80%.

Is there a way to fix this?

Rosetta will only run on CPU.
Einstein will run on CPU or GPU AND ... the GPU version requires 1 full CPU.

Running Einstein on ONLY the GPU plus the GPU required 1 CPU ... will likely be more than 20%.

I would suggest you change the Einstein preferences to only run GPU jobs on your machine.

The idle CPUs will then request Rosetta jobs.

Thanks for the suggestion, but I don't want to run anything on the GPU.

I am running only 2 projects: Rosetta at 80% and Einstein at 20%.

For Rosetta, BOINC manager keeps "Requesting new tasks for Intel GPU," which ALWAYS results in "Scheduler request completed: got 0 new tasks."

Then Einstein comes along a few minutes later and scores new tasks by "Requesting new tasks for CPU."

This results in my queue filling up with Einstein tasks, even though I'm running Rosetta at 80%.

Is there a way to fix this?

Rosetta will only run on CPU.
Einstein will run on CPU or GPU AND ... the GPU version requires 1 full CPU.

Running Einstein on ONLY the GPU plus the GPU required 1 CPU ... will likely be more than 20%.

I would suggest you change the Einstein preferences to only run GPU jobs on your machine.

The idle CPUs will then request Rosetta jobs.

Thanks for the suggestion, but I don't want to run anything on the GPU.

I have had to play with the BOINC CLIENT settings to get exactly what I want but these XML controls are pretty flexible.
https://boinc.berkeley.edu/wiki/Client_configuration

Try inserting the following TEXT file named "app_config.xml" in the Einstein Project directory. Change the "3" to the maximum Einstein jobs you want to run.


Tell BOINC to read it ....
OPTIONS: READ CONFIG FILE option

Look at the EVENT LOG and you should see a message that says it READ the config file.

------------------------------------ cut

<app_config>
<!--
I am a comment
-->
<project_max_concurrent> 3 </project_max_concurrent>

</app_config>

Thanks. Just for mental serenity, is this what I should see in the new Event Log?

6/7/2017 10:26:39 AM | rosetta@home | Config: excluded GPU. Type: intel_gpu. App: all. Device: all

If that is the type of GPU that you have on this machine, yes. It might also say

6/4/2017 7:10:45 AM | rosetta@home | Config: excluded GPU. Type: NVIDIA. App: all. Device: all

---

Rosetta Moderator: Mod.Sense

I am running only 2 projects: Rosetta at 80% and Einstein at 20%.

For Rosetta, BOINC manager keeps "Requesting new tasks for Intel GPU," which ALWAYS results in "Scheduler request completed: got 0 new tasks."

Then Einstein comes along a few minutes later and scores new tasks by "Requesting new tasks for CPU."

This results in my queue filling up with Einstein tasks, even though I'm running Rosetta at 80%.

Is there a way to fix this?

Rosetta will only run on CPU.
Einstein will run on CPU or GPU AND ... the GPU version requires 1 full CPU.

Running Einstein on ONLY the GPU plus the GPU required 1 CPU ... will likely be more than 20%.

I would suggest you change the Einstein preferences to only run GPU jobs on your machine.

The idle CPUs will then request Rosetta jobs.

Thanks for the suggestion, but I don't want to run anything on the GPU.

I have had to play with the BOINC CLIENT settings to get exactly what I want but these XML controls are pretty flexible.
https://boinc.berkeley.edu/wiki/Client_configuration

Try inserting the following TEXT file named "app_config.xml" in the Einstein Project directory. Change the "3" to the maximum Einstein jobs you want to run.


Tell BOINC to read it ....
OPTIONS: READ CONFIG FILE option

Look at the EVENT LOG and you should see a message that says it READ the config file.

------------------------------------ cut

<app_config>
<!--
I am a comment
-->
<project_max_concurrent> 3 </project_max_concurrent>

</app_config>

Thanks. I'm not sure it will fix the problem I have with the queue filling up with Einstein tasks, but it may keep Einstein from using all of my (8) cores when those tasks become critical.

BTW, the Event Log says "Found app_config.xml," for Einstein, not that it read it. Is this ok or did I do something wrong?

I am running only 2 projects: Rosetta at 80% and Einstein at 20%.

For Rosetta, BOINC manager keeps "Requesting new tasks for Intel GPU," which ALWAYS results in "Scheduler request completed: got 0 new tasks."

Then Einstein comes along a few minutes later and scores new tasks by "Requesting new tasks for CPU."

This results in my queue filling up with Einstein tasks, even though I'm running Rosetta at 80%.

Is there a way to fix this?

Rosetta will only run on CPU.
Einstein will run on CPU or GPU AND ... the GPU version requires 1 full CPU.

Running Einstein on ONLY the GPU plus the GPU required 1 CPU ... will likely be more than 20%.

I would suggest you change the Einstein preferences to only run GPU jobs on your machine.

The idle CPUs will then request Rosetta jobs.

Thanks for the suggestion, but I don't want to run anything on the GPU.

I have had to play with the BOINC CLIENT settings to get exactly what I want but these XML controls are pretty flexible.
https://boinc.berkeley.edu/wiki/Client_configuration

Try inserting the following TEXT file named "app_config.xml" in the Einstein Project directory. Change the "3" to the maximum Einstein jobs you want to run.


Tell BOINC to read it ....
OPTIONS: READ CONFIG FILE option

Look at the EVENT LOG and you should see a message that says it READ the config file.

------------------------------------ cut

<app_config>
<!--
I am a comment
-->
<project_max_concurrent> 3 </project_max_concurrent>

</app_config>

Thanks. I'm not sure it will fix the problem I have with the queue filling up with Einstein tasks, but it may keep Einstein from using all of my (8) cores when those tasks become critical.

BTW, the Event Log says "Found app_config.xml," for Einstein, not that it read it. Is this ok or did I do something wrong?

Seems OK.
The XML commands that guide BOINC operation are pretty flexible.
I "think" that BOINC asks each project you are working on for work to fill the COMPUTING PREFERENCES "amount of work" you have set. Einstein just says "Here ... take all you want", but has nearer deadlines than Rosetta. It just takes awhile for the pipeline of work to stabilize to allow Rosetta jobs to get closer to deadline and resources dedicated to Rosetta to ramp up.


If you would like to get more EVENT LOG comments to play around, you might add the XML app line to your Einstein file. Einstein will say something like (an old sample message from event log):

Your app_config.xml file refers to an unknown application 'TEST'. Known applications: 'hsgamma_FGRP2', 'einstein_S6BucketLVE', 'einsteinbinary_BRP5', 'hsgamma_FGRP3', 'einstein_S6CasA', 'einsteinbinary_BRP4G', 'einsteinbinary_BRP6'



<app> <name> TEST </name> <max_concurrent> 6 </max_concurrent> </app>


This granularity was useful to me when running a project with lots of applications and I wanted to control the number of "resource hog" apps running at one time.

Now that things have settled down after the server upgrade, I thought I would report on the effectiveness of suggestions made here.

Thanks to Mod.Sense and rjs5. I have added the xml files they recommended.

I also made one further change to my settings: I changed Store at least 2.0 days of work and Store up to an additional 0.5 days of work to 2.3 and 0.2 respectively. I reasoned that having a smaller "additional" amount would result in fewer tasks added each time BOINC Manager determined it was time to download some Einstein tasks. It is now usually downloading only one task at a time.

Bottom line: Einstein tasks now make up roughly 20% of the tasks in my queue - success!

---