trace-agent is not using the configured log_level

[renchap]

In my /etc/datadog-agent/datadog.yml I am setting the log_level like this:
log_level: warning

This is correctly working for datadog-agent, but trace-agent is not using it and still outputs INFO level logs, even if reading the config file:

Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (main.go:89) - pid '21161' written to pid file '/opt/datadog-agent/run/trace-agent.pid'
Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (agent.go:225) - Loaded configuration: /etc/datadog-agent/datadog.yaml
Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (trace_writer.go:49) - Trace writer initializing with config: {MaxSpansPerPayload:1000 FlushPeriod:5s Upda
Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (stats_writer.go:42) - Stats writer initializing with config: {MaxEntriesPerPayload:12000 UpdateInfoPeriod
Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (service_writer.go:31) - Service writer initializing with config: {FlushPeriod:5s UpdateInfoPeriod:1m0s Se
Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (main.go:156) - trace-agent running on host tg-staging2.ovh.talegraph.net
Nov 06 22:25:55 tg-staging2 trace-agent[21161]: 2018-11-06 22:25:55 INFO (api.go:140) - listening for traces at http://localhost:8126
Nov 06 22:26:01 tg-staging2 trace-agent[21161]: 2018-11-06 22:26:01 INFO (service_writer.go:76) - flushed service payload to the API, time:483.180167ms, size:77 bytes
Nov 06 22:26:01 tg-staging2 trace-agent[21161]: 2018-11-06 22:26:01 INFO (trace_writer.go:97) - flushed trace payload to the API, time:484.055105ms, size:6557 bytes
Nov 06 22:26:05 tg-staging2 trace-agent[21161]: 2018-11-06 22:26:05 INFO (api.go:324) - [lang:ruby lang_version:2.5.1 interpreter:ruby-x86_64-linux tracer_version:0.16.1] -> trace
Nov 06 22:26:05 tg-staging2 trace-agent[21161]: 2018-11-06 22:26:05 INFO (trace_writer.go:97) - flushed trace payload to the API, time:358.4442ms, size:4855 bytes
Nov 06 22:26:10 tg-staging2 trace-agent[21161]: 2018-11-06 22:26:10 INFO (trace_writer.go:97) - flushed trace payload to the API, time:326.552803ms, size:4736 bytes

Environment

I am using the Chef cookbook to install Datadog on Debian.

# /opt/datadog-agent/embedded/bin/trace-agent --version
Version: 6.6.0
Git hash: 24d4c76
Git branch: HEAD
Build date: 2018-10-25 15:27:32.839833398 +0000 UTC m=+0.002630095
Go Version: 1.10.2

[gbbr]

This works fine, but there is a small problem here, the correct logging level to set is warn, not warning. This works in the rest of the agent because there is a hook for this, but that's just luck. Can you please try the same with warn and see if that works for you?

[gbbr]

Created #518 to fix this for 6.7.0

[renchap]

I confirm using warn is working correctly. Thanks!

[gbbr]

Glad to hear! Thanks for confirming. I will close this issue once my PR gets merged then. In the next version (6.7.0) we should hopefully be consistent with the main agent and support both "warn" and "warning".

[renchap]

Might I suggest to have a way to override it for the trace agent, for example an optional apm_log_level? I can open a new issue for this if you want :)

[gbbr]

I'm not sure if there is any value in adding that. I think it would just increase confusion. Besides the issue you've already opened (#517), which is a valid claim that we should fix, I think there is no need for a new option.