apache/flink · error · RuntimeException

Could not load the AvroSerializer class. You may be missing

Error message

Could not load the AvroSerializer class. You may be missing the 'flink-avro' dependency.

What it means

Thrown by DefaultAvroUtils.createAvroSerializer() when the type system asks for an AvroSerializer but flink-avro is absent (DefaultAvroUtils is the no-Avro fallback). The real AvroSerializer lives in flink-avro and cannot be loaded, so creating a serializer for an Avro type fails immediately instead of producing a broken Kryo serializer.

Source

Thrown at flink-core/src/main/java/org/apache/flink/api/java/typeutils/AvroUtils.java:144

                                + "'. "
                                + "You may be missing the 'flink-avro' dependency.");
            }
        }

        @SuppressWarnings({"rawtypes", "unchecked"})
        @Override
        public void addAvroGenericDataArrayRegistration(
                LinkedHashMap<String, KryoRegistration> kryoRegistrations) {
            kryoRegistrations.put(
                    AVRO_GENERIC_DATA_ARRAY,
                    new KryoRegistration(
                            Serializers.DummyAvroRegisteredClass.class,
                            (Class) Serializers.DummyAvroKryoSerializerClass.class));
        }

        @Override
        public <T> TypeSerializer<T> createAvroSerializer(Class<T> type) {
            throw new RuntimeException(
                    "Could not load the AvroSerializer class. "
                            + "You may be missing the 'flink-avro' dependency.");
        }

        @Override
        public <T> TypeInformation<T> createAvroTypeInfo(Class<T> type) {
            throw new RuntimeException(
                    "Could not load the AvroTypeInfo class. "
                            + "You may be missing the 'flink-avro' dependency.");
        }
    }
}

View on GitHub (pinned to 2f3c205e92)

Solutions

1. Add flink-avro to the job dependencies (provided scope for cluster deployment).
2. Confirm the jar is present at runtime in the TaskManager classpath (lib/ or user-jar).
3. If Avro use is accidental, switch the type to a non-Avro serializable type.

Example fix

// before: keyed state on Avro record, no flink-avro
ListState<MyAvroRecord> st = ctx.getListState(descriptor);

// after: ensure flink-avro is on the classpath, scope provided
mvn dependency:tree | grep avro  # then add flink-avro provided

boolean canCreateAvroSerializer = AvroUtils.tryGetAvroUtils().isPresent();
if (!canCreateAvroSerializer) {
    throw new IllegalStateException("flink-avro missing; cannot build AvroSerializer for keyed state");
}

When it happens

Trigger: createSerializer() on a TypeInformation backed by an Avro type, or explicit AvroSerializer instantiation through AvroUtils when flink-avro is not on the classpath.

Common situations: State backend / checkpoint requesting a serializer for a keyed state whose type is an Avro record, but the job was submitted without flink-avro; using Avro POJOs as keys without the format jar.

Related errors

• Could not load class 'org.apache.flink.formats.avro.utils.Av
• Unexpected serializer type.
• Could not load the AvroTypeInfo class. You may be missing th
• Could not find required Avro dependency.
• Could not create writer state serializer.

AI-assisted analysis of apache/flink@2f3c205e92 (2026-08-14). Data as JSON: /api/errors/6caba04c6c99b447. Report an issue: GitHub.