Scheduler change with 3.6.5? Schedule for every minute is not working anymore

[eh-gh]

Hi here,

I'm using a scheduler entry to check every minute a water level sensor. If the level has reached I switch on a water pump connected to another pin.
It does not work after upgrade from 3.6.4 to 3.6.5 (and with the latest dev release either).

Evidence is attached.
Can you check plz?

tnx
eh

emsesp_customizations.json
emsesp_schedule.json
latest_dev_log.txt
working_3.6.4_log.txt

ps.
the mqtt error in dev log can not be seen in log 3.6.5, but the scheduler does not work with that version either.

[MichaelDvP]

Seems you are running out of memory. Please attach the support information.

[eh-gh]

here you are
fyi: the version 3.6.5 does not show the "out of memory" but the scheduling problem exists.

tnx

emsesp_system_info.txt

[MichaelDvP]

Looks ok. Are you sure the command is not executed? We have changed the log level for the success message to debug.

EMS-ESP32/src/web/WebSchedulerService.cpp

Line 349 in 43777f4

EMSESP::logger().debug("Scheduled command '%s' with data '%s' was successful", cmd, data);

so you do not see it.

[proddy]

@eh-gh set your log level in the WebUI to ALL in System->Log and see if the Scheduler command is actually being called.

[eh-gh]

just tested with the some water: as you can see the water level has reached but the pump does not start
in case I would fall back to 3.6.4 the scheduler works.

[proddy]

I prefer to keep the nested and not split topics if possible. There must be another way, is it the publish_ha_sensor_config() ?

[eh-gh]

just verified the original problem with 3.7.0-dev.10
it works again
thanks for the amazing fast correction

[MichaelDvP]

shall I make a 3.6.6 patch?

I don't think this feature is used often, a patch is nice if it's not too much affort.

There must be another way,

We can limit the number of HA configs send at once. After start the telegrams come in one by one and HA config is created on each mqtt publish a small number of new entities. My log with counter and 10 sec nested mqtt publish shows:

2024-05-05 11:09:15.966 I 15: [emsesp] Recognized new boiler with deviceID 0x08
2024-05-05 11:09:16.033 I 16: [emsesp] Created 5 HA-boiler-configs
2024-05-05 11:09:19.911 I 17: [emsesp] Recognized new controller with deviceID 0x09
2024-05-05 11:09:20.208 I 18: [emsesp] Recognized new connect with deviceID 0x0B
2024-05-05 11:09:20.726 I 19: [emsesp] Recognized new thermostat with deviceID 0x10
2024-05-05 11:09:21.154 I 20: [emsesp] Recognized new mixer with deviceID 0x21
2024-05-05 11:09:21.571 I 21: [emsesp] Recognized new solar with deviceID 0x30
2024-05-05 11:09:26.002 I 22: [emsesp] Created 62 HA-boiler-configs
2024-05-05 11:09:26.325 I 23: [emsesp] Created 14 HA-solar-configs
2024-05-05 11:09:36.058 I 26: [emsesp] Created 45 HA-thermostat-configs
2024-05-05 11:09:46.059 I 27: [emsesp] Created 4 HA-thermostat-configs
2024-05-05 11:09:46.237 I 28: [emsesp] Created 8 HA-solar-configs
2024-05-05 11:09:46.494 I 29: [emsesp] Created 2 HA-mixer-configs
2024-05-05 11:09:56.003 I 31: [emsesp] Created 10 HA-boiler-configs
2024-05-05 11:09:56.239 I 32: [emsesp] Created 3 HA-thermostat-configs
2024-05-05 11:09:56.519 I 33: [emsesp] Created 4 HA-mixer-configs
2024-05-05 11:10:06.062 I 34: [emsesp] Created 7 HA-thermostat-configs
2024-05-05 11:10:16.001 I 35: [emsesp] Created 3 HA-boiler-configs

eh-gh have 77 boiler entities including dhw, heatpumps have over 100 without dhw, With esp without psram they will see this memory message also with older versions and boiler_ww mqtt.
As mentioned it is no issue, on the memory limit mqtt stops generating HA messages and continue on next publish, no config message is lost.

[MichaelDvP]

Oh, we have 2 different memory checks, we should sync them.

EMS-ESP32/src/emsdevice.cpp

Line 1771 in 8084b91

if (Mqtt::publish_ha_sensor_config(dv, name(), brand_to_char(), false, create_device_config)) {

is the check for publish, if fail the errormessage is created here:

EMS-ESP32/src/mqtt.cpp

Lines 611 to 617 in 8084b91

	if (heap_caps_get_free_size(MALLOC_CAP_8BIT) < 60 * 1024) { // checks free Heap+PSRAM
	if (operation == Operation::PUBLISH) {
	mqtt_message_id_++;
	mqtt_publish_fails_++;
	}
	LOG_WARNING("%s failed: low memory", operation == Operation::PUBLISH ? "Publish" : operation == Operation::SUBSCRIBE ? "Subscribe" : "Unsubscribe");
	return false; // quit

but the HA ggeneration also should break a bit before, but with different test:

EMS-ESP32/src/emsdevice.cpp

Line 1777 in 8084b91

if (ESP.getMaxAllocHeap() < (6 * 1024) || (!EMSESP::system_.PSram() && ESP.getFreeHeap() < (65 * 1024))) {

I think we should always check the heap_caps with 60k limit in mqtt and 65k or 70k in emsdevice. So we can always add a large boiler_data message without getting to the mqtt limit and create the remaining HA configs later.

[proddy]

thanks @MichaelDvP for fixing the HA memory issue