Lost LOGs - increase LOG backend queue size.

[Dusan Daniska]

Hello, how can I increase LOG backend queue size? I was comparing some sdk_config settings with some examples, but did not find difference and the example does not fire this message. I am not logging much more info than the example.

Thank you.

[Jakub Rzeszutko]

Hi,

How do you know you are loosing logs?

Do you use CLI with Logger or only Logger module?

[Dusan Daniska]

I am using CLI with Logger and I can see this in the terminal:

And also I can not see some logs that I am expecting.

[Jakub Rzeszutko]

Hi Dusan,

I assume you are referring to nrf_cli example.

In the main.c you need to update define: CLI_EXAMPLE_LOG_QUEUE_SIZE used in the macro NRF_CLI_DEF. By default it has value == 4.

You can always run command: cli_stats to check queue usage.

[Dusan Daniska]

I am initializing the NRF_CLI using this piece of code:

NRF_CLI_CMD_REGISTER(set, NULL, "help string test", cli_cmd_set);
NRF_CLI_UART_DEF(m_cli_uart_transport, 0, 64, 16);
NRF_CLI_DEF(m_cli_uart, "sequoia: ", &m_cli_uart_transport.transport, '\r', 4);

When I increase the "4" in the NRF_CLI_DEF to e.g. "5", the logs start to behave strangely, this is the output:

So even the error log is not complete...

If I increase the value to something greater, e.g. "12", I get this output:

-> I do not see the red error but I can see only first char of the "cli_prefix".

[Jakub Rzeszutko]

Hi,

I have few additional questions:
1. Do you process nrf_cli in main like in the example?

2. May you try to increase buffers in NRF_CLI_UART_DEF: 64 ->1024 and 16->1024? Does it change anything?

3. Can you share your project? It will be hard to find the problem otherwise.

[Dusan Daniska]

I have tested several combinations and have found out following: the problematic code is probably the NRF_LOG_HEXDUMP_INFO.

When using this NRF_LOG_HEXDUMP_INFO, I can not increase the queue size and I get the "lost logs" error. If I increase the queue size, the program fails somewhere (I can't see the leds blinking which indicates normal operation).

When I commented ou the NRF_LOG_HEXDUMP_INFO, I could increase the queue size without having the program crash and I do not have the "lost logs" error.

I did not need  

Any idea why? Could this some bug in the NRF_LOG_HEXDUMP_INFO?