Page 1 of 1
GPUS disabled
Posted: Sat Sep 19, 2020 3:56 am
by chabgood
Ubuntu 20, 7 GPU, 4 - 2080, 3 - 2080 supers.
I downloaded the 7.6.17 client. I rebooted and all 7 GPU were disabled. I ended up deleted all 7 GPU, then i rebooted, then all 7 GPUS showed up. Then a few days later i rebooted and then all 7 GPU's were disabled, ugggh, I deleted all GPU's then i rebooted and then all 7 GPU's showed up.
Any idea why this is happening after rebooting?
Re: GPUS disabled
Posted: Sat Sep 19, 2020 5:50 am
by Whompithian
On CentOS 8, I have had consistent problems with the latest beta client failing to recognize my GPUs
if there is a GPU section in the config file. If there are no GPU sections defined, it detects the GPUs properly, then writes the appropriate GPU sections to the config, which causes the client to fail to recognize the GPUs the next time it starts.
To work around this problem, I use "ExecStartPre" in the systemd unit file to copy a version of the config file that has no GPU section defined. The relevant section of `FAHClient.service`:
Code: Select all
[Service]
ExecStartPre=cp /usr/local/etc/fahclient/config.xml /etc/fahclient/config.xml
ExecStart=/usr/bin/FAHClient --config=/etc/fahclient/config.xml --chdir=/var/lib/fahclient/
Another option is to `chown -R root:root /etc/fahclient` or to `chattr i /etc/fahclient/config.xml` to prevent modification of the config file by the client. This approach, however, results in a lot of spam in the logs regarding insufficient permission to move the config file for backup. I have not tested disabling config backups to solve this.
Re: GPUS disabled
Posted: Sat Sep 19, 2020 7:28 am
by bruce
One possibility: If your GPUs have a high failure rate, FAH wants to stop you draining WUs that can be completed by somebody else. I will disable your resources.
Re: GPUS disabled
Posted: Sat Sep 19, 2020 7:36 am
by foldy
I would recommend to stay with public FAHclient 7.6.13. The beta versions are still buggy
Re: GPUS disabled
Posted: Sat Sep 19, 2020 11:32 am
by PantherX
Please note that the beta version which you downloaded (V7.6.17) wasn't meant to be a Beta version. It was a developmental version uploaded for the development team to test. It clearly has massive issues and a new version will be released when ready. We have raised it up with the development team that a different method to distribute a test build instead of using a Beta page. Let's wait and see what happens.