mqtt_connect(&client) does not return any value during operation, and the program is stuck in mqtt_connect(&client) and does not come out

Hi,

There's several configuration(s) and function calls missing from the partial code that you posted, so I am unable to test your scenario. Do you see the same issue with for instance mqtt_simple sample?

Kind regards,

Håkon

Call mt_mqtt_connect() in this function and find that the function has no return value. The program stops in this function. According to the log print information, the program stops in the function err = mqtt_connect(&client).

Hi,

I am very glad to receive your reply. Unfortunately, we have such a problem, because our customers are currently running devices using ncs1.3.0 and modem v1.2.3, and ncs1.3.0 does not support modem upgrades , So according to what you said, the modem needs to be upgraded to 1.2.7. This method cannot be achieved. Is there any way to solve this problem without upgrading the modem?

Hi,

Another note, the SDK I currently test is 1.5.1, and the nrf_modem under this version is v1.1.0, so updating nrf_modem 1.0.3 alone cannot solve this problem.

The ncs v1.5.1 tagged version uses nrfxlib v1.5.1, which holds this as the newest version:

https://developer.nordicsemi.com/nRF_Connect_SDK/doc/1.5.1/nrfxlib/nrf_modem/doc/CHANGELOG.html

You might have manually updated the "nrfxlib" to something else?

Do you close the socket before re-trying to connect to the mqtt broker?

Kind regards,

Håkon 

Hi,

I did not close the socket . I used the mfw_nrf9160_1.2.7 firmware in ncs1.6.0 (nrf_modem library has been upgraded to v1.0.3). After a day of testing, the problem still reappeared, but it did not solve the problem. Encountered a new bug, please take a look at what is going on, thank you!

Hi,

george.zhou said:

I did not close the socket

You should close the socket, via the mqtt_disconnect() or mqtt_abort(). If not, you risk leaving the old socket still open after the mqtt_connect() call.

george.zhou said:

After a day of testing, the problem still reappeared, but it did not solve the problem. Encountered a new bug, please take a look at what is going on, thank you!

Which new bug did you observe?

Kind regards,

Håkon

Hi,

george.zhou said:

I did not close the socket

You should close the socket, via the mqtt_disconnect() or mqtt_abort(). If not, you risk leaving the old socket still open after the mqtt_connect() call.

george.zhou said:

After a day of testing, the problem still reappeared, but it did not solve the problem. Encountered a new bug, please take a look at what is going on, thank you!

Which new bug did you observe?

Kind regards,

Håkon

Tested according to the method you said, and did not solve the problem, indicating that the problem I encountered is different from the problem you mentioned, and it should be a new problem.

First of all, the connection to mqtt is blocked, and I did not reconnect to mqtt because the program is stuck in the mqtt connection and does not return, but once the mqtt connection is successful, I will use mqtt_disconnect() to close the socket after sending the data.I can make sure that the socket is closed every time before connecting mqtt

Hi,

Could you try setting these configurations to see if the scenario improves?

CONFIG_NRF_MODEM_LIB_SENDMSG_BUF_SIZE=2048
CONFIG_NRF_MODEM_LIB_HEAP_SIZE=2048

Kind regards,

Håkon

Hi，

I added the two macros you mentioned and tested them for a day on ncs1.6.0 and mfw_nrf9160_1.2.7, and found that the problem was improved. Some get stuck for 3 and a half minutes and return an error code, and sometimes get stuck for more than ten minutes and return an error code. The error returned is EINVAL. I will test for a few more days to see if the problem is completely solved. If it does solve the problem, I will test again on ncs1.5.1 and mfw_nrf9160_1.2.3 to see if this problem can also be solved. I will reply to the results after the test to you, thank you so much!

Hi，

I tested one more day, unfortunately, the problem reappeared, stuck in mqtt connection with no return value all the time.

Hi,

george.zhou said:

I added the two macros you mentioned and tested them for a day on ncs1.6.0 and mfw_nrf9160_1.2.7, and found that the problem was improved. Some get stuck for 3 and a half minutes and return an error code, and sometimes get stuck for more than ten minutes and return an error code. The error returned is EINVAL.

This is expected behavior if the nRF is unable to connect, due to network connectivity problems.

george.zhou said:

I tested one more day, unfortunately, the problem reappeared, stuck in mqtt connection with no return value all the time.

And this was with which SDK and modem fw?

The connect() call never returned?

Kind regards,

Håkon