MongoDB profiling

When running MongoDB in Docker, diagnostics logs are sent to STDOUT by default. Since MongoDB version 4.4 these logs have been in a structured JSON format, which makes it a little cumbersome to use the profiling tool in mtools.

Use the Docker command to export Docker container logs (including the MongoDB ones) and jq to sanitize all non JSON logs (from potential other processes in the container):

docker logs mongo_container_id \
 | jq -R "fromjson? | . " -c \
 > mongo_diagnostics.json.log

Notice: The mongo_container_id can be found using docker container ls.

Convert the sparkly “new” 4.4 JSON format log to the old legacy format using convert-json-logs-to-legacy-tool, because the log analyzer doesn’t support JSON:

generate_mplot_logs.py --log mongo_diagnostics.json.log \
> mongo_diagnostics.log

Now analyze the logs using mtools:

mloginfo --queries mongo_diagnostics.log

I don’t have a ton of profiling experience myself… looking forward to see where this will take me.