apache/flink · error · UnsupportedOperationException

Could not find required Avro dependency.

Error message

Could not find required Avro dependency.

What it means

Serializers.DummyAvroKryoSerializerClass is a placeholder Kryo serializer registered for Avro types purely to keep savepoint registration IDs identical to pre-1.4 Flink, where flink-core shipped Avro serializers. Since Avro is no longer on flink-core's classpath, the dummy's write() always throws UnsupportedOperationException 'Could not find required Avro dependency.' — hitting it means a write actually tried to serialize an Avro object through the dummy.

Source

Thrown at flink-core/src/main/java/org/apache/flink/api/java/typeutils/runtime/kryo/Serializers.java:167

    /**
     * This is used in case we don't have Avro on the classpath. Flink versions before 1.4 always
     * registered special Serializers for Kryo but starting with Flink 1.4 we don't have Avro on the
     * classpath by default anymore. We still have to retain the same registered Serializers for
     * backwards compatibility of savepoints.
     */
    public static class DummyAvroRegisteredClass {}

    /**
     * This is used in case we don't have Avro on the classpath. Flink versions before 1.4 always
     * registered special Serializers for Kryo but starting with Flink 1.4 we don't have Avro on the
     * classpath by default anymore. We still have to retain the same registered Serializers for
     * backwards compatibility of savepoints.
     */
    public static class DummyAvroKryoSerializerClass<T> extends Serializer<T> {
        @Override
        public void write(Kryo kryo, Output output, Object o) {
            throw new UnsupportedOperationException("Could not find required Avro dependency.");
        }

        @Override
        public T read(Kryo kryo, Input input, Class<? extends T> aClass) {
            throw new UnsupportedOperationException("Could not find required Avro dependency.");
        }
    }
}

View on GitHub (pinned to 2f3c205e92)

Solutions

1. Add flink-avro (and the matching Avro version) to the job's dependencies so real Avro serializers replace the dummy.
2. Explicitly register Avro classes/serializers: Serializers.registerAvroSerializers(env) or env.registerTypeWithKryoSerializer(MyAvroRecord.class, AvroSerializer).
3. Better: use AvroTypeInfo (SpecificRecord base class) instead of generic Kryo serialization for Avro types.

Example fix

// before (pom.xml has no avro dependency; Kryo falls back to dummy)
DataStream<MyRecord> s = env.fromElements(record);

// after (pom.xml)
<dependency>
  <groupId>org.apache.flink</groupId>
  <artifactId>flink-avro</artifactId>
  <version>${flink.version}</version>
</dependency>

static boolean avroAvailable() {
    try {
        Class.forName("org.apache.avro.specific.SpecificRecordBase");
        return true;
    } catch (ClassNotFoundException e) {
        return false;
    }
}
// if (!avroAvailable()) fail fast with a clear 'add flink-avro' message

When it happens

Trigger: Serializing an Avro record (org.apache.avro.specific.*) via Kryo in a job that does NOT include flink-avro (or avro dependency) in the user jar; restoring legacy state whose Kryo registration maps Avro classes to the dummy; relying on default Kryo fallback for Avro types instead of registering flink-avro's serializers.

Common situations: Migrating a pre-1.4 job that serialized Avro objects with Kryo; forgetting to add flink-avro / avro to the pom causing the dummy registration to be used at runtime.

Related errors

• Could not load class 'org.apache.flink.formats.avro.utils.Av
• Could not load the AvroSerializer class. You may be missing
• Could not load the AvroTypeInfo class. You may be missing th
• Error during Java serialization.
• Error during Java deserialization.

AI-assisted analysis of apache/flink@2f3c205e92 (2026-08-14). Data as JSON: /api/errors/3d71ba6dd635f872. Report an issue: GitHub.