Memory leak in container

[jwatzk]

What's the issue?

I see a memory leak running b2500-meter in a container using a sligthly modified vzlogger meter.

Your setup

• Installation method: Docker
• Storage system: VenusE V3, Firmware 144
• Power meter source: vzlogger

Configuration

[GENERAL]
# Comma-separated list of device types to emulate (ct001, shellypro3em, shellyemg3, shellyproem50)
DEVICE_TYPE = shellypro3em

[VZLOGGER]
IP = 192.168.186.120
PORT = 8181
UUID = 2171e0fa-0752-11eb-adc1-0242ac120003
POWER_CALCULATE = True
POWER_INPUT_UUID = 2171e0fa-0752-11eb-adc1-0242ac120003
POWER_OUTPUT_UUID = 65ee0920-0752-11eb-adc1-0242ac120003
## Per-powermeter throttling override (optional)
THROTTLE_INTERVAL = 1

Error messages or logs

!### Start of container:
CONTAINER ID   NAME                          CPU %     MEM USAGE / LIMIT     MEM %     NET I/O   BLOCK I/O   PIDS
f43eb4391026   b2500meter-vz_b2500-meter_1   1.18%     38.09MiB / 3.641GiB   1.02%     0B / 0B   26MB / 0B   11

!### after ~5 days running:
CONTAINER ID   NAME                          CPU %     MEM USAGE / LIMIT     MEM %     NET I/O   BLOCK I/O        PIDS
f43eb4391026   b2500meter-vz_b2500-meter_1   2.69%     1.311GiB / 3.641GiB   35.99%    0B / 0B   24.5MB / 4.9MB   11

Additional info

I don't think my modified forked code is causing this, but don't know how to troubleshoot this inside the container since there are no system/python tools installed in the container.
Maybe someone observed this with the original vzlogger meter?

[jwatzk]

Thanks for your investigation. Indeed the size of the container logs corresponds with the memory consumption. The messages flooding the log are from the print() function in throttling.py every second:

55:                print(
56:                    f"Throttling: Waiting {wait_time:.1f}s before fetching fresh values..."
57:                )

I would suggest to replace all the print() with logger.debug()

[jwatzk]

With changed code i reduced logging in the container to a few bytes, but still the container memory increases ~4MB per hour.
Can this be normal behaviour?

[tomquist]

Would've surprised me if the logging change fixed the memory leak. Logs don't go into memory. My suspicion is more that it has to do with throttling. Can you try with throttling disabled?

[jwatzk]

Great, throtteling disabled fixed the memory increase.

[jwatzk]

Any idea why throtteling causes the memory increase? I discoverd that i can not disable throtteling because of VenusE's control loop is oscillating, see below example setting throtteling from 0 to 5 at ~16:02