Sensor does connect to the formula after disconnection

[PierreRustOrange]

Test scenario:

• run the sensor
• start the formula :
  • it receives the report and correctly generates power reports
• stop the formula , restart it
  • the formula does not receives any report from the sensor

The following command were used when testing:

Sensor:

docker run --privileged --rm --name sensor --network="host"  powerapi/hwpc-sensor \
         -v /sys:/sys  \
         -v /var/lib/docker/containers:/var/lib/docker/containers:ro     \
           -n sensor \
           -f 2000 \
           -r socket -U 127.0.0.1 -P 12000 \
           -s "rapl" -o -e "RAPL_ENERGY_PKG" \
           -s "msr"     -e "TSC" -e "APERF" -e "MPERF" \
           -c "core"    -e "CPU_CLK_THREAD_UNHALTED:REF_P" \
                        -e "CPU_CLK_THREAD_UNHALTED:THREAD_P" \
                        -e "LLC_MISSES"\
                    -e "INSTRUCTIONS_RETIRED"

BTW, the sensor in the latest docker image does not returns its own version : I: 21-10-26 15:28:23 build: version undefined (rev: undefined) (Sep 28 2021 - 14:40:24)

Formula:

python -m smartwatts --debug --config-file config_file.json

version: today's pull on master.

configuration file:

{
  "verbose": true,
  "stream": true,
  "input": {
    "puller": {
      "model": "HWPCReport",
      "type": "socket",
      "uri": "127.0.0.1",
      "port": 12000
    }
  },
  "output": {
    "pusher_power": {
      "type": "csv",
      "uri": ".",
      "model" : "PowerReport"
    }
  },
  "cpu-frequency-base": 2300,
  "cpu-frequency-min": 800,
  "cpu-frequency-max": 5100,
  "cpu-error-threshold": 2.0,
  "disable-dram-formula": true,
  "sensor-report-sampling-interval": 1000
}

[PierreRustOrange]

I had a quick look at this issue but actually I can't fond whats causing the sensor to stop when the socket is closed at the server side : In reporting_actor, errors when writing out reports seems to be silently ignored and I don't see any code reacting to connection loss. Obviously I'm missing something here, any idea on where I should look ? Thanks !

[PierreRustOrange]

I've finally found the root cause, when writing to the closed socket we get an EPIPE signal which not handled and stops the sensor. We should probably ignore this signal and handle the error manually by reconnecting the socket.