Quote from: papampi on November 17, 2017, 08:05:22 PM

Quote from: Stubo on November 17, 2017, 06:58:06 PM

Quote from: leenoox on November 17, 2017, 06:17:05 PM

I believe they use $UTILIZATIONS and $GPU_COUNT to show it in telegram message. They need to fetch that info from the log or alternatively query it from within the telegram script itself.

If we switch all logs to reside on the tmpfs it shouldn't be problem logging those if we rotate/flush the logs periodicaly.

As for the logging from 3main, that should definetely be removed.

I just checked on telegram (had not thought of that) and it looks to be doing mostly it's own gathering of data for telegrams of type bali and damNmad as opposed to log mining. However, there is a singular instance of tailing the 5_restartlog in papampi telegram to get GPU_UTILIZATION. That is an easy switch to use the same one liner as bali and damNmad use to get GPU_UTILIZATIONS.

Will change it for next version for sure

In the mean time any one wants can change it with this line :

Code:

GPU_UTILIZATIONS=$(nvidia-smi --query-gpu=utilization.gpu --format=csv,noheader,nounits | tr '\n' ' ')

Excellent. The tweaks to the logging that I recommend (specifically the 3main ones as above) have already paid off for me. I was able to identify a possible OC problem with a GPU in one of my rigs by looking at a nice clean screenlog.0 file. Looking at it previously with the output from both the watchdog and temp control scripts was confusing as heck and I had been chasing that miner restart/hang problem for weeks.