Normally

INFO

is fine for our use case, but we have seen on occasion something goes wrong and the job pod consumes all the free memory on the kubernetes node and we would like to capture additional logs when we’ve detected abnormal memory usage.

In my testing it seemed like once a flow was running changing the log level didn’t have any impact.

I have this code in my flow:

@task()
def log_debug():
    for i in range(10):
        logger = prefect.context.get("logger")
        logger.debug("Test Debug Message")
        sleep(30)