9.10. Avro Schema Registry Converter

The Avro Schema Registry converter handles data written by Apache Avro using a Confluent Schema Registry. The schema registry is a centralized store of versioned Avro schemas.

To use the Avro converter, specify type = "avro-schema-registry" in your converter definition.

Note that Confluent requires Avro 1.8 and the Confluent client JARs, which are not bundled with GeoMesa.

9.10.1. Configuration

The Avro Schema Registry converter supports parsing Avro data using a Confluent schema registry. To configure the schema registry set schema-registry = "<URL of schema registry>" in your converter definition.

The Avro record being parsed is available to field transforms as $1.

The Avro Schema Registry Converter is an extension of the Avro Converter, therefore the Avro Transform Functions can be used to extract fields out of the parsed Avro record.

9.10.2. Example Usage

For this example we’ll assume the following Avro schema is registered in the schema registry as version 1:

{
  "namespace": "org.locationtech",
  "type": "record",
  "name": "SchemaRegistryMessageV1",
  "fields": [
    {
      "name": "lat",
      "type": "Double"
    },
    {
      "name": "lon",
      "type": "Double"
    }
  ]
}

We’ll also assume the following Avro schema is registered in the schema registry as version 2:

{
  "namespace": "org.locationtech",
  "type": "record",
  "name": "SchemaRegistryMessageV2",
  "fields": [
    {
      "name": "lat",
      "type": "Double"
    },
    {
      "name": "lon",
      "type": "Double"
    },
    {
      "name": "extra",
      "type": "String"
    }
  ]
}

Below is a sample Avro record encoded using schema version 1:

{
  "lat": 45.0,
  "lon": 45.0
}

Here’s a sample Avro record encoded using schema version 2:

{
  "lat": 45.0,
  "lon": 45.0,
  "extra": "Extra Test Field"
}

Let’s say we want to convert our Avro records into simple features. We notice that between the two schema versions there are 3 attributes:

  • lat
  • lon
  • extra

The following converter config would be sufficient to parse the Avro records that have been encoded using multiple schema version defined in the schema registry:

{
  type        =     "avro-schema-registry"
  schema-registry = "http://localhost:8080"
  sft         =     "testsft"
  id-field    =     "uuid()"
  fields = [
    { name = "lat",    transform = "avroPath($1, '/lat')" },
    { name = "lon",    transform = "avroPath($1, '/lon')" },
    { name = "extra",  transform = "avroPath($1, '/extra')",
    { name = "geom",   transform = "point($lon, $lat)" }
  ]
}

Note that in the simple feature, the extra field will be null for Avro records encoded using schema version 1 and will be populated for records encoded using schema version 2.