Log config warnings in the logs, not on stderr

[alexbozhenko]

Currently, config parsing warnings are printed on stderr.

# nats-server -c test.conf  --log log.out
test.conf:9:1: invalid use of field "ping_interval": ping_interval should be converted to a duration

Depending on how the server is deployed(systemd, k8s, nix, etc...) and how the logs are aggregated, this stderr may be lost and will not show up where all the other nats servers logs are.

Test plan:

This will make print the config parsing warnings using whatever logging we configured, not on stderr, where it can be never noticed.

# ./nats-server -c test.conf  --log log.out
# cat log.out
[95618] 2025/12/05 14:49:44.456825 [INF] Starting nats-server
[95618] 2025/12/05 14:49:44.456893 [INF]   Version:  2.14.0-dev
...
[95618] 2025/12/05 14:49:44.456932 [INF] Using configuration file: test.conf (sha256:d20faaf4b973de79fbfcc43c0c4784365473a235a04d681c8a6df05e5781ba91)
[95618] 2025/12/05 14:49:44.456941 [WRN] Configuration warning: test.conf:9:1: invalid use of field "ping_interval": ping_interval should be converted to a duration
...
[95618] 2025/12/05 14:49:44.459155 [INF] Listening for client connections on 0.0.0.0:4222
[95618] 2025/12/05 14:49:44.459963 [INF] Server is ready

Signed-off-by: Alex Bozhenko [email protected]

[neilalexander]

My understanding is that we always write these configuration errors directly to stderr as we don't know at that point whether we have successfully parsed any other log-related options that could redirect them elsewhere or obscure them from view.

Do you have any thoughts @derekcollison?

(To my knowledge Kubernetes and systemd both capture stdout+stderr by default, can't speak for Nix though.)

[derekcollison]

IIRC we do stderr for errors and warnings and stdout for INFO. But I would need to look, that was done over a decade ago.