ncreplayer 1.0.2

Playback netCDF CF DSG files onto a Kafka topic

ncreplayer

Small little utility designed to load a CF DSG compliant netCDF file and replay it back onto a Kafka topic in either batch mode or stream mode. Optionally control the timestamps and timedeltas in the original file using configuration parameters.

Data is formatted as described in the AVRO schema file schema.avsc. You can choose to serialize the data as avro, msgpack or the default json.

WHY?

This utility is designed to make my life easier. It isn't inteded to be used by many people but has been useful for:

• Playing back data in a netCDF file with different times and intervals than what is defined in the file.
• Setting up a quick stream of data for load testing.
• Transitioning parts of large systems that rely on static files to stream processing.
• Accepting a standardized data format from collaborators (netCDF/CF/DSG) and being able to stream it into larger systems.

Data Format

Simplest example

{
  "uid": "1",
  "time": "2019-04-01T00:00:00Z",
  "lat": 30.5,
  "lon": -76.5,
}

Full Example

{
  "uid": "1",
  "gid": null,
  "time": "2019-04-01T00:00:00Z",
  "lat": 30.5,
  "lon": -76.5,
  "z": null,
  "values": {
    "salinity": 30.2,
    "temperature":  46.5
  },
  "meta": ""
}

• values objects are optional and are a multi-type AVRO map.
• meta is optional and open-ended. It is intended to carry along metadata to describe the values. I recommend using nco-json. This is useful if the system listening to these messages needs some context about the data, like streaming data to a website. YMMV.

Configuration

This program uses Click for the CLI interface. I probably spent 50% of the time making this utility just playing around with the CLI interface. I have no idea if what I came up with is mind-blowingly awesome or a bunch of crap. It works. Comments are welcome.

$ ncreplay
Usage: ncreplay [OPTIONS] FILENAME COMMAND [ARGS]...

Options:
  --brokers TEXT                 Kafka broker string (comman separated)
                                 [required]
  --topic TEXT                   Kafka topic to send the data to  [required]
  --packing [json|avro|msgpack]  The data packing algorithm to use
  --registry TEXT                URL to a Schema Registry if avro packing is
                                 requested
  --uid TEXT                     Variable name, global attribute, or value to
                                 use for the uid values
  --gid TEXT                     Variable name, global attribute, or value to
                                 use for the gid values
  --meta / --no-meta             Include the `nco-json` metadata in each
                                 message?
  --help                         Show this message and exit.

Commands:
  batch   Batch process a netCDF file in chunks, pausing every [chunk]...
  stream  Streams each unique timestep in the netCDF file every [delta]...

batch

$ ncreplay /path/to/file.nc batch --help
Usage: ncreplay batch [OPTIONS]

  Batch process a netCDF file in chunks, pausing every [chunk] records
  for [delta] seconds. Optionally change the [starting] time of the file
  and/or change each timedelta using the [factor] and [offset] parameters.

Options:
  -s, --starting [%Y-%m-%d|%Y-%m-%dT%H:%M:%S|%Y-%m-%d %H:%M:%S]
  -f, --factor FLOAT
  -o, --offset FLOAT
  -d, --delta FLOAT
  -c, --chunk INTEGER
  --help                          Show this message and exit.

stream

$ ncreplay /path/to/file.nc stream --help
Usage: ncreplay stream [OPTIONS]

  Streams each unqiue timestep in the netCDF file every [delta] seconds.
  Optionally you can control the [starting] point of the file and this will
  re-calculate all of the timestamps to match the original timedeltas.

Options:
  -s, --starting [%Y-%m-%d|%Y-%m-%dT%H:%M:%S|%Y-%m-%d %H:%M:%S]
  -d, --delta FLOAT
  --help                          Show this message and exit

Development / Testing

There are no tests (yet) but you can play around with the options using files included in this repository

First setup a Kafka ecosystem

$ docker run -d --net=host \
    -e ZK_PORT=50000 \
    -e BROKER_PORT=4001 \
    -e REGISTRY_PORT=4002 \
    -e REST_PORT=4003 \
    -e CONNECT_PORT=4004 \
    -e WEB_PORT=4005 \
    -e RUNTESTS=0 \
    -e DISABLE=elastic,hbase \
    -e DISABLE_JMX=1 \
    -e RUNTESTS=0 \
    -e FORWARDLOGS=0 \
    -e SAMPLEDATA=0 \
    --name ncreplayer-testing \
  landoop/fast-data-dev:1.0.1

Then setup a listener

$ docker run -it --rm --net=host \
  landoop/fast-data-dev:1.0.1  \
    kafka-console-consumer \
      --bootstrap-server localhost:4001 \
      --topic axds-ncreplayer-data

Now batch or stream a file:

# Batch
$ ncreplay tests/data/gda_example.nc batch -d 10 -c 10

# Stream
$ ncreplay tests/data/gda_example.nc stream -d 10

To test the avro packing, setup a listener that will unpack the data automatically:

$ docker run -it --rm --net=host \
  landoop/fast-data-dev:1.0.1  \
    kafka-avro-console-consumer \
      --bootstrap-server localhost:4001 \
      --property schema.registry.url=http://localhost:4002 \
      --topic axds-ncreplayer-data

And use avro packing

$ ncreplay --packing avro tests/data/gda_example.nc batch -d 10 -c 10