Hazelcast and python there is no suitable de-serializer for type -120

Question

hello i guess have problem with client and member config which config should i use as you can see i am inserting json as data when i call get_data it returns with no problem but when i try to use predicate-sql it gives me error "hazelcast.errors.HazelcastSerializationError: Exception from server: com.hazelcast.nio.serialization.HazelcastSerializationException: There is no suitable de-serializer for type -120. This exception is likely caused by differences in t he serialization configuration between members or between clients and members."

@app.route('/insert_data/<database_name>/<collection_name>', methods=['POST'])
def insert_data(database_name, collection_name):
    client = hazelcast.HazelcastClient(cluster_members=[
        url
    ])
    dbname_map = client.get_map(f"{database_name}-{collection_name}").blocking()

    if request.json:
        received_json_data = request.json
        received_id = received_json_data["_id"]
        del received_json_data["_id"]
        dbname_map.put(received_id, received_json_data)
        client.shutdown()
        return jsonify()

    else:
        client.shutdown()
        abort(400)

---

@app.route('/get_data/<database_name>/<collection_name>', methods=['GET'])
def get_all_data(database_name, collection_name):
    client = hazelcast.HazelcastClient(cluster_members=[
        url
    ])
    dbname_map = client.get_map(f"{database_name}-{collection_name}").blocking()
    entry_set = dbname_map.entry_set()
    output = dict()
    datas = []
    for key, value in entry_set:
        value['_id'] = key
        output = value
        datas.append(output)

    client.shutdown()
    return jsonify({"Result":datas})

---

@bp.route('/get_query/<database_name>/<collection_name>/<name>', methods=['GET'])
def get_query_result(database_name, collection_name,name):
     client = hazelcast.HazelcastClient(cluster_members=[
        url
    ])
    predicate_map = client.get_map(f"{database_name}-{collection_name}").blocking()
    predicate = and_(sql(f"name like {name}%"))
    entry_set = predicate_map.values(predicate)
    #entry_set = predicate_map.entry_set(predicate)

    send_all_data = ""
    for x in entry_set:
        send_all_data += x.to_string()
        send_all_data += "\n"
        print(send_all_data)
    # print("Retrieved %s values whose age is less than 30." % len(result))
    # print("Entry is", result[0].to_string())
    # value=predicate_map.get(70)
    # print(value)
    return jsonify()

i try to change hazelcast.xml according to hazelcast-full-example.xml but i can't start hazelcast after the changes and do i really have to use serialization ? hazelcast version:4.1 python:3.9

Answer 1

This is most likely happening because you are putting entries of the type dictionary to the map, which is serialized by the pickle because you didn't specify a serializer for that and the client does not know how to handle that correctly, so it fallbacks to the default serializer. However, since pickle serialization is Python-specific, servers cannot deserialize it and throw such an exception.

There are possible solutions to that, see the https://hazelcast.readthedocs.io/en/stable/serialization.html chapter for details.

I think the most appropriate solution for your use case would be Portable serialization which does not require a configuration change or code on the server-side. See the https://hazelcast.readthedocs.io/en/stable/serialization.html#portable-serialization

BTW, client objects are quite heavyweight, so you shouldn't be creating them on demand like this. You can construct it once in your application and share and use it in your endpoints or business-logic code freely since it is thread-safe. The same applies to the map proxy you get from the client. It can also be re-used.