Fahclient won't upload WU without me.

[Fold Mining]

If I'm logged out, this error repeats over and over:

Code: Select all

20:15:30:WARNING:WU01:FS00:Exception: Could not get IP address for assign3.stanford.edu: Name or service not known
20:15:30:ERROR:WU01:FS00:Exception: Could not get an assignment
20:15:31:WU00:FS00:Sending unit results: id:00 state:SEND error:NO_ERROR project:8616 run:396 clone:0 gen:51 core:0xa4 unit:0x000000380002894c57b796be35ee4124
20:15:31:WU00:FS00:Uploading 4.00MiB to 155.247.166.220
20:15:31:WU00:FS00:Connecting to 155.247.166.220:8080
20:15:31:WARNING:WU00:FS00:WorkServer connection failed on port 8080 trying 80
20:15:31:WU00:FS00:Connecting to 155.247.166.220:80
20:15:31:WARNING:WU00:FS00:Exception: Failed to send results to work server: Failed to connect to 155.247.166.220:80: Network is unreachable
20:15:31:WU00:FS00:Trying to send results to collection server
20:15:31:WU00:FS00:Uploading 4.00MiB to 155.247.166.219
20:15:31:WU00:FS00:Connecting to 155.247.166.219:8080
20:15:31:WARNING:WU00:FS00:WorkServer connection failed on port 8080 trying 80
20:15:31:WU00:FS00:Connecting to 155.247.166.219:80
20:15:31:ERROR:WU00:FS00:Exception: Failed to connect to 155.247.166.219:80: Network is unreachable

After I log back in, Fahclient uploads the WU at its next attempt.

I'm using an i5 and GeForce GT 730, but the Fahclient says AMD64 and GPUs: 0.

Code: Select all

19:30:06:******************************** Build ********************************
19:30:06:    Version: 7.4.4
19:30:06:       Date: Mar 4 2014
19:30:06:       Time: 12:01:17
19:30:06:    SVN Rev: 4130
19:30:06:     Branch: fah/trunk/client
19:30:06:   Compiler: GNU 4.1.2 20080704 (Red Hat 4.1.2-46)
19:30:06:    Options: -std=gnu++98 -O3 -funroll-loops -mfpmath=sse -ffast-math
19:30:06:             -fno-unsafe-math-optimizations -msse2
19:30:06:   Platform: linux2 2.6.18-164.11.1.el5
19:30:06:       Bits: 64
19:30:06:       Mode: Release
19:30:06:******************************* System ********************************
19:30:06:        CPU: Intel(R) Core(TM) i5-4570 CPU @ 3.20GHz
19:30:06:     CPU ID: GenuineIntel Family 6 Model 60 Stepping 3
19:30:06:       CPUs: 4
19:30:06:     Memory: 31.34GiB
19:30:06:Free Memory: 31.03GiB
19:30:06:    Threads: POSIX_THREADS
19:30:06: OS Version: 4.4
19:30:06:Has Battery: false
19:30:06: On Battery: false
19:30:06: UTC Offset: -7
19:30:06:        PID: 1114
19:30:06:        CWD: /var/lib/fahclient
19:30:06:         OS: Linux 4.4.126-48-default x86_64
19:30:06:    OS Arch: AMD64
19:30:06:       GPUs: 0
19:30:06:       CUDA: Not detected
19:30:06:***********************************************************************

What am I doing wrong here?

[SteveWillis]

It looks like you are losing your internet connection. If you were running on windows I'd suggest going into the device manager and in your network adapters,properties,power management and uncheck the box next to Allow the computer to turn off this device to save power.
I'm sure there is a Linux equivalent but I haven't yet found where you can set it.

[Joe_H]

I'm using an i5 and GeForce GT 730, but the Fahclient says AMD64 and GPUs: 0.

AMD64 refers to the 64-bit extensions made by AMD to the 32-bit x86 architecture. Intel ended up having to adopt it when their planned 64-bit CPU architecture and processors failed to gain commercial acceptance in the PC market 15 or so years ago. So AMD64 is correct even if you have an Intel i5.

As for detecting the GPU, for that to work the nVidia proprietary drivers and OpenCL support packages need to be installed and active before installing the client. You did not post enough of the beginning of the log to show the folding client configuration information, but if it says GPUS - false, that has to be changed to true. Restarting the client should scan and detect the GPU.

[Fold Mining]

Sounds like I need to uninstall and reinstall the fah software. I had the nVidia proprietary drivers installed, but not the OpenCL rpms. Do I need to install anything besides beignet and opencl-headers?

How do I make sure the OpenCL packages are active after installation?

[bruce]

In my experience, when you install the proprietary NVidia drivers, the CUDA and OpenCl drivers are included. The two lines:
19:30:06: GPUs: 0
19:30:06: CUDA: Not detected
are a clear indication that the drivers needed by FAH have not been installed.

It's true that it's possible for a 3rd party to create a package including ONLY the video drivers, but I'd look around for other similar packages that do include the essential computational packages.
Some people have gone so far as to install the NVidia developer's package, which certainly includes those things, but you get other things that you won't really use.

[SteveWillis]

I don't know if your gpu is nvidia but if it is, this is how I install the drivers and it always works. The only time I tried a manual install from a file it broke Linux.


Installation of nvidia drivers into Linux Mint
pause folding. Some people finish all slots but I haven't found that to be necessary
sudo add-apt-repository ppa:graphics-drivers/ppa #Just the first time
sudo add-apt-repository ppa:xorg-edgers/ppa # Just the first time
sudo apt-get update
sudo apt-get install nvidia-xxx # xxx=driver version

Then go into driver manager in system settings, select the nouveau version, apply changes, DO NOT BOOT, select the new version and apply changes, then re-boot. This may possiblly not be necessary for you but if I don't do it I just get all folding slots failed.

[bruce]

@Steve:

The switching to nouveau and back is essential and I've crashed a few Linux installations by doing it wrong. When you update from version xxx to yyy do you need to switch to nouveau BEFORE the ... install nvidia-yyy step or AFTER?

[Fold Mining]

In my experience, when you install the proprietary NVidia drivers, the CUDA and OpenCl drivers are included. The two lines:
19:30:06: GPUs: 0
19:30:06: CUDA: Not detected
are a clear indication that the drivers needed by FAH have not been installed.

The nVidia drivers are installed including all of the OpenCl and CUDA libs:

Code: Select all

nvidia-computeG04 - NVIDIA driver for computing with GPGPU

/etc/OpenCL
/etc/OpenCL/vendors
/etc/OpenCL/vendors/nvidia.icd
/usr/bin/nvidia-cuda-mps-control
/usr/bin/nvidia-cuda-mps-server
/usr/bin/nvidia-modprobe
/usr/bin/nvidia-smi
/usr/lib/libOpenCL.so
/usr/lib/libOpenCL.so.1
/usr/lib/libOpenCL.so.1.0.0
/usr/lib/libcuda.so
/usr/lib/libcuda.so.1
/usr/lib/libcuda.so.390.48
/usr/lib/libnvidia-fatbinaryloader.so.390.48
/usr/lib/libnvidia-ml.so.1
/usr/lib/libnvidia-ml.so.390.48
/usr/lib/libnvidia-opencl.so.1
/usr/lib/libnvidia-opencl.so.390.48
/usr/lib/libnvidia-ptxjitcompiler.so.1
/usr/lib/libnvidia-ptxjitcompiler.so.390.48
/usr/lib64/libOpenCL.so
/usr/lib64/libOpenCL.so.1
/usr/lib64/libOpenCL.so.1.0.0
/usr/lib64/libcuda.so
/usr/lib64/libcuda.so.1
/usr/lib64/libcuda.so.390.48
/usr/lib64/libnvidia-fatbinaryloader.so.390.48
/usr/lib64/libnvidia-ml.so
/usr/lib64/libnvidia-ml.so.1
/usr/lib64/libnvidia-ml.so.390.48
/usr/lib64/libnvidia-opencl.so.1
/usr/lib64/libnvidia-opencl.so.390.48
/usr/lib64/libnvidia-ptxjitcompiler.so.1
/usr/lib64/libnvidia-ptxjitcompiler.so.390.48
/usr/share/man/man1/nvidia-cuda-mps-control.1.gz

They were there before I even tried setting up FAH. Nouveau is even blacklisted in modprobe.d
So what am I missing?

[SteveWillis]

@Steve:

The switching to nouveau and back is essential and I've crashed a few Linux installations by doing it wrong. When you update from version xxx to yyy do you need to switch to nouveau BEFORE the ... install nvidia-yyy step or AFTER?

I always do it after.

[bruce]

The nVidia drivers are installed including all of the OpenCl and CUDA libs:

Code: Select all

nvidia-computeG04 - NVIDIA driver for computing with GPGPU

/etc/OpenCL
/etc/OpenCL/vendors
/etc/OpenCL/vendors/nvidia.icd
/usr/bin/nvidia-cuda-mps-control
/usr/bin/nvidia-cuda-mps-server
/usr/bin/nvidia-modprobe
/usr/bin/nvidia-smi
/usr/lib/libOpenCL.so
/usr/lib/libOpenCL.so.1
/usr/lib/libOpenCL.so.1.0.0
/usr/lib/libcuda.so
/usr/lib/libcuda.so.1
/usr/lib/libcuda.so.390.48
/usr/lib/libnvidia-fatbinaryloader.so.390.48
/usr/lib/libnvidia-ml.so.1
/usr/lib/libnvidia-ml.so.390.48
/usr/lib/libnvidia-opencl.so.1
/usr/lib/libnvidia-opencl.so.390.48
/usr/lib/libnvidia-ptxjitcompiler.so.1
/usr/lib/libnvidia-ptxjitcompiler.so.390.48
/usr/lib64/libOpenCL.so
/usr/lib64/libOpenCL.so.1
/usr/lib64/libOpenCL.so.1.0.0
/usr/lib64/libcuda.so
/usr/lib64/libcuda.so.1
/usr/lib64/libcuda.so.390.48
/usr/lib64/libnvidia-fatbinaryloader.so.390.48
/usr/lib64/libnvidia-ml.so
/usr/lib64/libnvidia-ml.so.1
/usr/lib64/libnvidia-ml.so.390.48
/usr/lib64/libnvidia-opencl.so.1
/usr/lib64/libnvidia-opencl.so.390.48
/usr/lib64/libnvidia-ptxjitcompiler.so.1
/usr/lib64/libnvidia-ptxjitcompiler.so.390.48
/usr/share/man/man1/nvidia-cuda-mps-control.1.gz

They were there before I even tried setting up FAH.

Maybe they're not in the path or wherever FAH expects to find them. ???


Why is CUDA not found? Nvidia's OpenCL depend on CUDA.

If you install the beta version of FAHClient it will give the version numbers of CUDA and OpenCL that it finds.... or you can try FAHBench which will allow you test each platform.

[SteveWillis]

By the way, Bruce, I'm folding fine and this line is in all of my log files:
2018-04-18:00:09:29: OpenCL: Not detected: Failed to open dynamic library 'libOpenCL.so':

[Fold Mining]

I went looking for fahbench at Stanford:

http://folding.stanford.edu/2013/03/19/fahbench-10

Vijay's link points to a 3rd world store for what I'm guessing from the photos is erectile dysfunction pills.

Where can I find the beta?

[bruce]

from here, I get a legitimate FAHBench 1.0 page. I conclude your DNS has been redirected to a bogus site.

By the way, FAHBench is now a v2.3 which I'm finding at http://fahbench.github.io/
Does that page look legitimate?

[Joe_H]

The two links to fahbench.com on that page for Dr' Pande's blog no longer work for me. As this is through a well locked down DNS here at my university, my assumption is that proteneer let the registration expire at some point. His blog site is also dead. I looked up the domain registration for fahbench.com, since 2016 it has been registered to someone in Indonesia.

As I recall, there was an area for some third party utilities for testing on the Stanford folding site, but I can not locate it now. An update link to download Fahbench was added there after the original site set up by proteneer went away.

[HayesK]

By the way, Bruce, I'm folding fine and this line is in all of my log files:
2018-04-18:00:09:29: OpenCL: Not detected: Failed to open dynamic library 'libOpenCL.so':

My linux setup includes a fix for the opencl index. Not sure if it will fix your issue, but has taken care of the gpu index issue on my hosts with Ubuntu 16.04.
# install opencl index detection
sudo apt install ocl-icd-opencl-dev --fix-missing