Folding@home Disconnected, Even After Reinstallation

[Coronablaster]

My computer is now displaying the following dialog:

--------

Folding@home Disconnected
The Folding@home software on your computer has most likely stopped.
Please restart it, then reload this page.

--------

Not knowing how to restart FAH, I simply deleted it, restarted the computer, and reinstalled a newer version (7.6.13, instead of the old 7.6.9). Nothing is changed! The new version displays the same message. Anyone know what might be wrong?

The computer is an Early 2008 MacPro3,1 dual processor with 22GB RAM, running Windows 10 Pro natively (that's not a mistake; there is a way to install it, and it runs beautifully; there is no emulator). The video card is an NVidia GeForce 8800GT, an older once without CUDA units, so, as is normal for this computer, that part of FAH is not running.

[Coronablaster]

UPDATE:

The Web-browser based control panel is set to run FAH at "Light Power, While I Am Working." The last time I used it, it was set to "Medium Power, Only When Idle." Upon beginning to use the computer again after a period when it was idle, the Resource Monitor displays:

FAHClient.exe Running

When I first touched the keyboard, Resource Monitor showed that the CPUs had been running at about half their capacity, and then dropped to nearly nothing when I started using the computer again, typical behavior for "Medium Power, Only When Idle." Two other FAH services had just terminated, and are now missing from Resource Monitor. I suspect that FAH is actually running correctly, but is not connecting to the control panel. If that sounds right, anyone know of a fix?

[Joe_H]

There was a bug in one of the last versions of the installer in getting to 7.6.13. That leaves the current installation looking for a password (not passkey) on the connection between the FAHControl and FAHClient running in the background. Normally a password is only needed for FAHcontrol to be used for connections to remote F@h clients on other systems.

Perhaps the easiest method to fix this that I know is reinstalling after uninstalling including data. First you may want to make a copy of the current config.xml file in the data directory to use for copying and pasting setup information such as username, team and passkey from.

Someone may know of another method that will fix this for you.

[Coronablaster]

UPDATE:

The two other FAH tasks that initially show in Resource Monitor, but that vanish when I start to use the computer, are:

FahCore_a7.exe
FAHCoreWrapper.exe

[Joe_H]

Those would only be active when your computer is doing CPU folding. When paused while waiting for "On Idle" they exit and are restarted when the computer does become idle.

[Coronablaster]

Thank you, Joe _H! Sounds as though folding is running normally, although the control panel can not connect. Reading the log.txt file:

17:10:57:WU02:FS00:0xa7:Completed 220000 out of 250000 steps (88%)

I can see that the current, long project is 88% complete. So, by reading the log, I should be able to wait to delete the data and re-install until after this fold completes, rather than now, which would lose the work so close to completion.

[Joe_H]

That sounds like a plan. Unfortunately the Web Control button that should allow setting the WU to "Finish" is not working. So the client can't be set to finish the current WU and not download the next. But deleting that new WU with the data before the reinstall is better than losing that much progress.

[Coronablaster]

After allowing the then current fold to finish, I deleted FAH and its data, and then reinstalled from a freshly downloaded installer, but something still seems to be wrong. The control panel now displays the running fold, but the controls work inconsistently. Set it to run "Only when idle," and it either (different results at different times) runs for a while at idle and then stops, or never runs. The log file displays some warnings, if they might explain what is wrong:

04:15:07:WU01:FS00:0xa7:Project: 16503 (Run 0, Clone 1961, Gen 50)
04:15:07:WU01:FS00:0xa7:Unit: 0x0000003b8f59f36f5eae574d20a499b2
04:15:07:WU01:FS00:0xa7:Digital signatures verified
04:15:07:WU01:FS00:0xa7:Calling: mdrun -s frame50.tpr -o frame50.trr -x frame50.xtc -cpi state.cpt -cpt 15 -nt 3
04:15:08:WU01:FS00:0xa7:Steps: first=12500000 total=250000
04:15:12:WU01:FS00:0xa7:Completed 160722 out of 250000 steps (64%)
04:15:18:FS00:Shutting core down
04:15:19:WU01:FS00:0xa7:WARNING:Console control signal 1 on PID 18516
04:15:19:WU01:FS00:0xa7:Exiting, please wait. . .
04:15:23:WU01:FS00:FahCore returned: INTERRUPTED (102 = 0x66)
04:16:05:WU01:FS00:Starting
04:16:05:WARNING:WU01:FS00:Changed SMP threads from 3 to 8 this can cause some work units to fail
04:16:05:WARNING:WU01:FS00:AS lowered CPUs from 8 to 3
04:16:05:WU01:FS00:Running FahCore: "C:\Program Files (x86)\FAHClient/FAHCoreWrapper.exe" C:\Users\jonsi\AppData\Roaming\FAHClient\cores/cores.foldingathome.org/v7/win/64bit/Core_a7.fah/FahCore_a7.exe -dir 01 -suffix 01 -version 706 -lifeline 12500 -checkpoint 15 -np 3
04:16:06:WU01:FS00:Started FahCore on PID 5972
04:16:06:WU01:FS00:Core PID:2508
04:16:06:WU01:FS00:FahCore 0xa7 started

To run FAH overnight, I set the control panel to "While I'm working" and full power. The CPU section of Resource Monitor shows the fold occurring at about the intensity of a typical fold at light power, which sounds about right, given the log file's statement that the fold is running on 3 CPUs, instead of 8. CPUs 1, 2, and 3 are running somewhat more intensely than the other five, but not even any of them are running beyond about 60% of capacity.

[Knish]

this kind of reminds me of the Chrome browser web control issue. If that is your browser, perhaps give a cache clearing a try?

[Joe_H]

The WU shown in the log was downloaded when your client was set to light, it will never use more than the 3 CPU threads it was set to at assignment time. Once finished the next WU would use more CPU threads.

Code: Select all

04:16:05:WARNING:WU01:FS00:Changed SMP threads from 3 to 8 this can cause some work units to fail
04:16:05:WARNING:WU01:FS00:AS lowered CPUs from 8 to 3

How the OS schedules processing of the CPU threads to individual processor cores will depend on the OS and the processor hardware.

Not certain what is going on with the Web Control for you, perhaps someone else has run into this problem or something similar.

[Coronablaster]

Browser is Firefox. Have cleared caches.

"The WU shown in the log was downloaded when your client was set to light, it will never use more than the 3 CPU threads it was set to at assignment time. Once finished the next WU would use more CPU threads."

Interesting. The control panel seems to suggest that one can control the amount of CPU used, but you're saying the maximum possible CPU usage is whatever was set when the WU was downloaded? So, download at full power, then lower if you please. Download at low power, no possibility to increase!

Joe_H, you've been patient and helpful with this, thank you.

[Joe_H]

Yes, the slider was designed at one point in the creation of the current client, it was a while later that they found an increased rate of failures related to CPU thread count. WUs would be downloaded at one setting, then a later increase in CPU threads assigned to that WU would cause it to fail. This is not an issue the vast majority of the time when reducing the thread count. Basically a project might have WUs of a size that could work on up to 4 threads, but will fail on more. So a thread count cap was added to the downloaded WUs, they can be decreased from that, then returned to that value, but not over that set number.

I do know they are working on developing a better method of having the number of CPU threads adjusted automatically to usable numbers. But no idea how long that may take to be released. The underlying Gromacs code for the CPU folding cores was originally targeted more at the researcher who could run a simulation on their own workstation or server and provide the various data and settings to be input. If something failed on an error they could adjust the settings and restart the process. For the F@h folding core they have to wrap that Gromacs code with additional software code to provide the inputs, capture the output, and do the appropriate action on receipt of various errors. They have a number of additional items that would be useful for the software to do automatically, getting it to for it right is the hard part.

[PantherX]

Welcome to the F@H Forum Coronablaster,

It's my understanding that you would like to fold on the CPU during idle while awake and Full overnight? If that's correct, this is what you can do:

Open up Advanced Control (AKA FAHControl)
Set the Folding Power to Full
Right-click the CPU Slot and select "On Idle" so it will only fold when the system is not in use
At nighttime, you can right-click the CPU Slot and untick "On Idle" which will cause it to start folding immediately
During daytime, enable the "On Idle" option again

BTW, CPU folding happens at a a very low priority so in most cases, it will not have an impact on other applications. Thus, you can leave your system folding 24/7 without issues as long as you're happy with the thermal, noise and electricity usage

[Coronablaster]

Joe_H, Why don't they just fudge the issue of the number of cores available to a WU? Have an optional setting that would always boost CPU use to "Full" temporarily when downloading WUs, then return it to wherever it had been before. I can't believe the WU take long to download. No one would even notice.

PantherX, I have been toggling between "Full" and "Light" via the Web Control Panel, which appears to yield a similar result. I had noticed that FAH seemed not to bother much of anything when set to full. I've only had trouble with busy Zoom sessions when bandwidth to the Internet was low, when it helped to kill FAH. It may be that Zoom needs to toggle its uses of the processor just a hair faster than FAH can allow? Makes no sense! To be honest, the main reason for setting to "Light" was the temperature in the room, and the surprising expense of running these old Xeon processors endlessly. It still seems to contribute a lot at "Light" CPU use.

FAH has been important to me, because it lets me feel I'm fighting back against the virus. Beats sitting at home, waiting for the world to change.

[Coronablaster]

PantherX, another option would be separate controls for CPU use. Allow "While I'm working" to be "Light," and "On Idle" to be "Full."