Haberdasher

Haberdasher is a simple command wrapper designed to consume log messages from its wrapped command's stderr stream and retransmit them someplace else. It also contains standard PID1 goodies for healthier container execution.

Simple demonstration

The foo.py program simply ticks off integers as log messages every 2 seconds.

$ ./haberdasher python3 foo.py
2020/09/14 16:03:00 Initializing haberdasher.
2020/09/14 16:03:00 Configured emitter: stderr
Python starting
{"ecs.version":"1.5.0","@timestamp":"2020-09-14T16:03:02.556065987-04:00","labels":{},"tags":[],"message":"0"}
{"ecs.version":"1.5.0","@timestamp":"2020-09-14T16:03:04.558082983-04:00","labels":{},"tags":[],"message":"1"}
{"ecs.version":"1.5.0","@timestamp":"2020-09-14T16:03:06.560023837-04:00","labels":{},"tags":[],"message":"2"}
^C2020/09/14 16:03:07 Signal received: interrupt
2020/09/14 16:03:07 Sending signal to 415770
2020/09/14 16:03:07 Trigering emitter shutdown

You can see that using the stderr emitter, it simply prints the received messages. Since the output of foo.py was unstructured, each log line that Haberdasher received is wrapped in a basic Elastic Common Schema message.

If Haberdasher receives a structured log message from its wrapped process, it leaves it alone and retransmits it unmodified.

$ ./haberdasher python3 foo.py --json
2020/09/14 16:05:02 Initializing haberdasher.
2020/09/14 16:05:02 Configured emitter: stderr
Python starting
{"i": 0}
{"i": 1}
{"i": 2}
^C2020/09/14 16:05:09 Signal received: interrupt
2020/09/14 16:05:09 Sending signal to 416367
2020/09/14 16:05:09 Trigering emitter shutdown

Configuring Haberdasher

Haberdasher is configured entirely from environment variables.

• HABERDASHER_EMITTER - configures the emitter to use. stderr is default, but kafka is also supported.
• HABERDASHER_TAGS - for unstructured log lines received, Haberdasher can add ECS tags to the wrapped messages. This value should be a serialized JSON list.
• HABERDASHER_LABELS - for unstructured log lines received, Haberdasher can add ECS labels to the wrapped messages. This value should be a serialized JSON object whose values are all strings.
• HABERDASHER_STDERR_PRETTY - if the stderr emitter is used, setting this to a non-empty string will result in the JSON being prettified before printing to stderr. This is useful in developer environments to make the messages easier to read.
• HABERDASHER_KAFKA_BOOTSTRAP - if the kafka emitter is used, this is required and points to the bootstrap listener for your Kafka cluster
• HABERDASHER_KAFKA_TOPIC - if the kafka emitter is used, this is required and names the Kafka topic log messages should be written to

Adding it to your Dockerfile

To use Haberdasher in a container, you only have to make two small modifications to your Dockerfile.

1. In a RUN stanza, include:

   curl -L -o /usr/bin/haberdasher https://github.com/RedHatInsights/haberdasher/releases/latest/download/haberdasher_linux_amd64 &&
   chmod 755 /usr/bin/haberdasher

2. Your ENTRYPOINT command should be: ["/usr/bin/haberdasher"]

And that's it! Rebuild and you're up and running.