XMR-STAK [CUDA] Error: the launch timed out and was terminated

Question

I'm using Ubuntu 16.04 and xmr-stak-portbin-linux.tar.gz

The miner when run manually with Nvidia GPU, runs without error, but gives error when run as a startup service.

Earlier, I was doing only CPU mining. I created a startup service xmr-service to run the miner at startup, which worked fine

[Unit]
Description=xmr-stak

[Service]
ExecStart=/usr/local/tmp/xmr-stak.sh -c "/usr/local/tmp/config.txt"
User=root

[Install]
WantedBy=multi-user.target
This was working fine at startup for CPU mining.

Now the system also had Nvidia Quadro K620 GPU but with no installed drivers. So I installed Nvidia drivers 384.81 and Cuda 9.0.176

Then I ran xmr-stak.sh manually. Following is the nvidia.txt that the miner generated.

"gpu_threads_conf" :
[
  // gpu: Quadro K620 architecture: 50
  //      memory: 1781/1995 MiB
  //      smx: 3
  { "index" : 0,
    "threads" : 90, "blocks" : 9,
    "bfactor" : 0, "bsleep" :  0,
    "affine_to_cpu" : false,
  },

],

The miner crashed, so I changed the parameters to following:

"gpu_threads_conf" :
[
  // gpu: Quadro K620 architecture: 50
  //      memory: 1781/1995 MiB
  //      smx: 3
  { "index" : 0,
    "threads" : 1, "blocks" : 1,
    "bfactor" : 8, "bsleep" :  25,
    "affine_to_cpu" : false,
  },

],

On again running the miner manually, the miner was running with these parameters.

Then I did

sudo systemctl daemon-reload
and restarted the system

On sudo systemctl status xmr-service , it shows following:

workstation xmr-stak.sh[795]: [2017-12-12 18:16:37] : Result accepted by the pool. 
workstation xmr-stak.sh[795]: [2017-12-12 18:16:39] : Result accepted by the pool. 
workstation xmr-stak.sh[795]: [2017-12-12 18:16:42] : Result accepted by the pool. 
workstation xmr-stak.sh[795]: [CUDA] Error gpu 0: </home/fireice/xmr-stak/xmrstak/backend/nvidia/nvcc_code/cuda_extra.cu>:244
workstation xmr-stak.sh[795]: terminate called after throwing an Instance of 'std::runttme_error'
workstation xmr-stak.sh[795]: what(): [CUDA] Error: the launch timed out and was terminated
workstation xmr-stak.sh[795]: /usr/local/tmp/xmr-stak.sh: line 3: 824 Aborted (core dumped) $SCRIPT_PATH/stak/ld-linux-x86-64.so.2 --library-path $SCRIPT_PATH/stak $SCRIPT_PATH/stak/xmr-stak $* 
workstation systemd[1]: xmr-stak.service: Main process exited, code=exited, status=134/n/a 
workstation systemd[1]: xmr-stak.service: Unit entered failed state. 
workstation systemd[1]: xmr-stak.service: Failed with result 'exit-code'. 

The miner when run manually with Nvidia GPU, runs without error, but gives error when run as a startup service. Also the startup service worked fine with only CPU.

What could be the problem?

Answer 1

I haven't played with my systemctl script in a while, but here are my notes. They are close to yours, except for my file name is xmr.service and there are two additional commands at the end. Hope this helps.

---

Create systemd startup script Create new file xmr.service in /etc/systemd/system

Add the following snippet:

[Unit]
Description=XMRig
After=network-online.target

[Service]
ExecStart=/PATH/TO/MINER

[Install]
WantedBy=multi-user.target

Perform these commands:

systemctl daemon-reload

systemctl start xmr.service

systemctl enable xmr.service