How to force worker to re-detect my hardware?

Benchmarking process description.

Benchmarks are run on all equipment, the first time the worker service is launched. The duration of the GPU redshift benchmark can go for a long time, up to two hours. Later when the vorker is restarted, only the benchmarks of the network will be re-launched. Other benchmarks are taken from the cache, because they are tied to specific equipment and are not updated when the worker is restarted.

The progress of benchmarking process can be checked by executing sudo docker ps command

It shows all the Docker containers that are currently running through your system.

Normal results for this command should look like this:

You can check resources availiable for worker after benchmarking: sonmcli worker devices

Normal results for this command should look like this:

If sudo docker ps gives empty results and worker do not responding to sonmlcli worker devices command, check your worker settings

How to relaunch benchmarks.

You'll need to relaunch benchmarks in case of:

  • Replacing/adding a new hardware component on Supplier's machine
  • Incorrect benchmark value/if Worker did not detected specific component
  • You have changed the hardware parameters (for example, GPU overclock).

WARNING! You need to correctly turn off your Worker instance before relaunching benchmarks.

To remove particular benchmark value you should know it's ID. Visit GitHub here to see benchmarks full list.

Use sonmcli worker benchmark remove <benchmark_id> to remove particular benchmark cashed value.

EXAMPLE: I overclocked my CPU. So, i need to restart cpu-sysbench-multi and cpu-sysbench-single benchmarks

I will execute sonmcli worker benchmark remove 0 and sonmcli worker benchmark remove 1 commands

You can also remove all cashed values of all benchmarks. Use sonmcli worker benchmark purge command for this.

Applying changes

Restart worker service: sudo service sonm-worker restart. Worker will run missing benchmarks, wait for benchmarking process to end, after that - disengage maintenance mode.

Check your resources availiable for worker to ensure that everything went ok: sonmlcli worker devices