Our correspondence helps with understanding Gobblin. ☺

We’re focusing on the source, so no writers are involved yet. We want to 
extract both the key and value from each kafka message.

There are three classes that extend abstract class, KafkaSource:
KafkaDeserializerSource,
KafkaSimpleSource and
UniversalKafkaSource.

Each has method, getExtractor(), that returns these respective instances:

KafkaDeserializerExtractor;
KafkaSimpleExtractor and
a class found by ClassAliasResolver. resolveClass.

The first two have method, decodeRecord(), that takes parameter, 
ByteArrayBasedKafkaRecord, and each calls ByteArrayBasedKafkaRecord 
.getMessageBytes(). Neither calls ByteArrayBasedKafkaRecord .getKeyBytes(), so 
these sources won’t work.

The UniversalKafkaSource class looks interesting because it will search for a 
class that extends KafkaExtractor and instantiate it. But there’s a catch: the 
class must be annotated with @Alias. If class KafkaSimpleJsonExtractor was 
annotated with an @Alias, I think I could use the UniversalKafaSource along 
with property, gobblin.source.kafka.extractorType. As you know, 
KafkaSimpleJsonExtractor. decodeRecord() combines both 
ByteArrayBasedKafkaRecord. getKeyBytes() and ByteArrayBasedKafkaRecord. 
getMessageBytes(), these result we seek.

Or did we miss something?

From: Vicky Kak [mailto:vicky....@gmail.com]
Sent: Friday, April 13, 2018 12:01 AM
To: user@gobblin.incubator.apache.org
Subject: Re: KafkaSimpleJsonExtractor

Are you not having this configuration?
***********************************************************************
writer.builder.class=org.apache.gobblin.writer.SimpleDataWriterBuilder
writer.file.path.type=tablename
writer.destination.type=HDFS
writer.output.format=txt
***********************************************************************

I scanned the code for the SimpleDataWriterBuilder which does uses the 
SimpleDataWriter that writes the byte array to the HDFS( I don't see the hdfs 
configuration so assume
that the file is only in hdfs format and it would be copied to hadoop file 
system after it is created).


It seems that you want to store the key/value in the Json format ( Or any other 
format) and not the way it is being done by the default SimpleDataWriter. You 
may look for
configuring the KafkaDeserializerExtrator with required deserialized type as 
Json( or other).

Once this is done then you can use KafkaSimpleJsonExtractor which was asked at 
the beginning.

So to conclude what I understand is that you don't want the data to be stored 
in the byte array, you require json (or other format).





On Thu, Apr 12, 2018 at 8:09 PM, Kidwell, Jack 
<jack_kidwel...@comcast.com<mailto:jack_kidwel...@comcast.com>> wrote:
Thank you for your time spent on our question.

This quick start document,
https://gobblin.readthedocs.io/en/latest/case-studies/Kafka-HDFS-Ingestion/,
shows 
“source.class=org.apache.gobblin.source.extractor.extract.kafka.KafkaSimpleSource”.
 That class overrides method, getExtractor, to return “new 
KafkaSimpleExtractor(state)”, and class, KafkaSimpleExtractor, overrides 
method, decodeRecord, to return just the Kafka message value.

  return kafkaConsumerRecord.getMessageBytes()

Since we want both key and value stored in HDFS, class, 
KafkaSimpleJsonExtractor, appears to provide the desired decodeRecord method:

protected byte[] decodeRecord(ByteArrayBasedKafkaRecord messageAndOffset) 
throws IOException {
        long offset = messageAndOffset.getOffset();

        byte[] keyBytes = messageAndOffset.getKeyBytes();
        String key = (keyBytes == null) ? "" : new String(keyBytes, CHARSET);

        byte[] payloadBytes = messageAndOffset.getMessageBytes();
        String payload = (payloadBytes == null) ? "" : new String(payloadBytes, 
CHARSET);

        KafkaRecord record = new KafkaRecord(offset, key, payload);

        byte[] decodedRecord = gson.toJson(record).getBytes(CHARSET);
        return decodedRecord;
    }

The architecture diagram leads us to think that the exactor is the point where 
both Kafka key and value are visible in the Gobblin pipeline:
https://gobblin.readthedocs.io/en/latest/Gobblin-Architecture/#gobblin-constructs

We are eager to learn from you.


From: Vicky Kak [mailto:vicky....@gmail.com<mailto:vicky....@gmail.com>]
Sent: Thursday, April 12, 2018 9:45 AM
To: user@gobblin.incubator.apache.org<mailto:user@gobblin.incubator.apache.org>
Subject: Re: KafkaSimpleJsonExtractor

For writing the data to the HDFS you need to take a look at the Writer 
Implementation, please look for it. The Extractor is used in pulling the data 
which is thereafter passed through various stages before it is stored via the 
Writer.
What I mean by the other subscriber was to add one more consumer which will  
read the key/value e.g kafka-console-consumer.sh which comes with the kafka.
I was not earlier aware that you wanted to store the data in HFDS as you had 
given the reference of the Extractor so it was not clear to me.
Please take a look at the Kafka to HDFS writer and see if that helps, in case 
it is not doing what is exactly required by you then you may have to plugin the 
customized writer.


On Thu, Apr 12, 2018 at 6:47 PM, Kidwell, Jack 
<jack_kidwel...@comcast.com<mailto:jack_kidwel...@comcast.com>> wrote:
We want to extract both key and value from a Kafka message and publish the 
combined information to HDFS. Keys contain numeric ids for the strings 
contained in values.

Please explain “other subscriber”. Are you proposing a different Kafka message 
structure?

KafkaSimpleJsonExtractor.java caught my attention because it extracts both keys 
and values and puts them in decodedRecord.

From: Vicky Kak [mailto:vicky....@gmail.com<mailto:vicky....@gmail.com>]
Sent: Thursday, April 12, 2018 1:17 AM
To: user@gobblin.incubator.apache.org<mailto:user@gobblin.incubator.apache.org>
Subject: Re: KafkaSimpleJsonExtractor

I am not sure what you are asking for. Do you want to see the keys/values 
rendered in the logs while the extraction is being done?
Why can't you have the other subscriber to the kafka which will render the 
values rather than KafkaExtractor implementation rendering the same?


On Wed, Apr 11, 2018 at 11:17 PM, Kidwell, Jack 
<jack_kidwel...@comcast.com<mailto:jack_kidwel...@comcast.com>> wrote:
Hi,

Using gobblin_0.10.0, we want to use module, KafkaSimpleJsonExtractor.java in 
order to see both kafka record keys and values.

How does one configure a job to achieve it?




Reply via email to