Can't load NVML library
cmxrootx opened this issue · comments
Hi,
I'm using Unraid, I updated to the new version and got this error:
20211025 09:25:39 WARN: Can't load NVML library, dlopen(25): failed to load libnvidia-ml.so, libnvidia-ml.so: cannot open shared object file: No such file or directory
How can i fix this ?
I receive that as a warning as well but the miner still works. Are you saying it shuts down?
I'm sorry I didn't read the complete log.
Yes, that one is a warning, but it fails after that one. And enters in a restart loop
Here is my log:
`20211025 18:01:26 T-Rex NVIDIA GPU miner v0.24.4 - [Linux]
20211025 18:01:26 r.1e965a96b3b9
20211025 18:01:26
20211025 18:01:26 WARN: Can't load NVML library, dlopen(25): failed to load libnvidia-ml.so, libnvidia-ml.so: cannot open shared object file: No such file or directory
20211025 18:01:26 WARN: NVML error, code 12
20211025 18:01:26 WARN: Can't initialize NVML. GPU monitoring will be disabled.
20211025 18:01:26
20211025 18:01:26 NVIDIA Driver version N/A
20211025 18:01:26
20211025 18:01:26 + GPU #0: [00:02.0|2489] GeForce RTX 3060 Ti, 7981 MB
20211025 18:01:26 + GPU #1: [00:03.0|2489] GeForce RTX 3060 Ti, 7982 MB
20211025 18:01:26 + GPU #2: [00:07.0|2489] GeForce RTX 3060 Ti, 7982 MB
20211025 18:01:26
20211025 18:01:26 WARN: DevFee 1% (ethash)
20211025 18:01:26
20211025 18:01:26 URL : stratum+tcp://daggerhashimoto.eu.nicehash.com:3353
20211025 18:01:26 USER: .
20211025 18:01:26 PASS: x
20211025 18:01:26 WRK : ***
20211025 18:01:26
20211025 18:01:26 WARN: NVML is disabled. You won't see GPUs stats.
20211025 18:01:26
20211025 18:01:26 Starting on: daggerhashimoto.eu.nicehash.com:3353
20211025 18:01:26 ApiServer: HTTP server started on 0.0.0.0:4067
20211025 18:01:26 ----------------------------------------------------
20211025 18:01:26 For control navigate to: http://172.17.0.1:4067/trex
20211025 18:01:26 ----------------------------------------------------
20211025 18:01:26 Using protocol: stratum2.
20211025 18:01:26 Extranonce is set to: 8abce5
20211025 18:01:26 Authorizing...
20211025 18:01:26 Authorized successfully.
20211025 18:01:26 ethash epoch: 449, diff: 1.02 G
20211025 18:01:26 TREX: Can't initialize device [ID=0, GPU #0], can't lock core clock, NVML wasn't initialized
20211025 18:01:26 WARN: Miner is going to shutdown...
20211025 18:01:27 Main loop finished. Cleaning up resources...
20211025 18:01:27 ApiServer: stopped listening on 0.0.0.0:4067
20211025 18:01:29 T-Rex finished.
20211025 18:01:30 WARN: WATCHDOG: T-Rex does not exist anymore, restarting...
20211025 18:01:31 T-Rex NVIDIA GPU miner v0.24.4 - [Linux]
20211025 18:01:31 r.1e965a96b3b9
20211025 18:01:31
20211025 18:01:31 WARN: Can't load NVML library, dlopen(25): failed to load libnvidia-ml.so, libnvidia-ml.so: cannot open shared object file: No such file or directory
20211025 18:01:31 WARN: NVML error, code 12
20211025 18:01:31 WARN: Can't initialize NVML. GPU monitoring will be disabled.
20211025 18:01:31
20211025 18:01:31 NVIDIA Driver version N/A`
Thanks.
What nvidia driver version are you running?
I'm running 470.74
Hmm, I'm running v495.29.05 . I wonder if that's an issue.
The NVML warnings (WARN: Can't load NVML library, dlopen(25): failed to load libnvidia-ml.so, libnvidia-ml.so: cannot open shared object file: No such file or directory) don't appear to be the problem.
The real issue is this line:
20211025 18:01:26 TREX: Can't initialize device [ID=0, GPU #0], can't lock core clock, NVML wasn't initialized
Do you use the mt or cclock flags at all in your config?
Yes I use lock_cclock "800" and mt is disabled.
--
I updated my drivers to match yours, and let cclock "0", And is working, but the WebUI is not reporting consumption data, I guess t-rex now requires NVML to report power consumption and things like that?
I found my way to fix the issue, installing libnvidia-ml-dev on the container using the console tool..
apt-get update && apt-get install libnvidia-ml-dev
Cheers.
Good find! I'll get that added to the container.