RE: nRF Cloud coap hangs after a few hours using nRF9160

SCollins — Tue, 23 Apr 2024 23:37:10 GMT

Hi Pete, thanks for your detailed response. We did consider a similar fix but I was unsure if it would cause any other problems with the coap client. Thank you very much for your help.

RE: nRF Cloud coap hangs after a few hours using nRF9160

Pete Skeggs — Tue, 23 Apr 2024 23:29:04 GMT

Hi Sam, this is a known issue in NCS 2.6.0. We are working on a fix.

The problem occurs for NON transfers when the response from the server never comes. The Zephyr coap_client library's coap_client_req() function handles NON transfer responses differently than CON transfers (no timeouts, no retries).

In that situation, in nrf_cloud_coap_transport.c: client_transfer(), this line waits forever:

(void)k_sem_take(&cb_sem, K_FOREVER); /* Wait for coap_client to exhaust retries */

As a workaround, change the line above to:

#define NON_RESP_WAIT_S 3 /* Wait for coap_client to exhaust retries when reliable transfer selected, * otherwise wait a finite time because response might never come. */ err = k_sem_take(&cb_sem, reliable ? K_FOREVER : K_SECONDS(NON_RESP_WAIT_S)); if (!err) { LOG_DBG("Got callback"); } else { LOG_DBG("Got timeout: %d", err); err = 0; /* Ignore, since caller selected non-reliable transfer */ }

RE: nRF Cloud coap hangs after a few hours using nRF9160

helsing — Wed, 17 Apr 2024 13:38:55 GMT

Hi Sam, just letting you know we have started looking into this.

Best regards,
Håkon