13.3. Kafka Connector Tutorial
Introduction
The Kafka Connector for Presto allows access to live topic data from Apache Kafka using Presto. This tutorial shows how to set up topics and how to create the topic description files that back Presto tables.
Installation
This tutorial assumes familiarity with Presto and a working local Presto installation (see Deploying Presto). It will focus on setting up Apache Kafka and integrating it with Presto.
Step 1: Install Apache Kafka
Download and extract Apache Kafka.
Note
This tutorial was tested with Apache Kafka 0.8.1. It should work with any 0.8.x version of Apache Kafka.
Start ZooKeeper and the Kafka server:
$ bin/zookeeper-server-start.sh config/zookeeper.properties
[2013-04-22 15:01:37,495] INFO Reading configuration from: config/zookeeper.properties (org.apache.zookeeper.server.quorum.QuorumPeerConfig)
...
$ bin/kafka-server-start.sh config/server.properties
[2013-04-22 15:01:47,028] INFO Verifying properties (kafka.utils.VerifiableProperties)
[2013-04-22 15:01:47,051] INFO Property socket.send.buffer.bytes is overridden to 1048576 (kafka.utils.VerifiableProperties)
...
This will start Zookeeper on port 2181 and Kafka on port 9092.
Step 2: Load data
Download the tpch-kafka loader from Maven central:
$ curl -o kafka-tpch https://repo1.maven.org/maven2/de/softwareforge/kafka_tpch_0811/1.0/kafka_tpch_0811-1.0.sh
$ chmod 755 kafka-tpch
Now run the kafka-tpch program to preload a number of topics with tpch data:
$ ./kafka-tpch load --brokers localhost:9092 --prefix tpch. --tpch-type tiny
2014-07-28T17:17:07.594-0700 INFO main io.airlift.log.Logging Logging to stderr
2014-07-28T17:17:07.623-0700 INFO main de.softwareforge.kafka.LoadCommand Processing tables: [customer, orders, lineitem, part, partsupp, supplier, nation, region]
2014-07-28T17:17:07.981-0700 INFO pool-1-thread-1 de.softwareforge.kafka.LoadCommand Loading table 'customer' into topic 'tpch.customer'...
2014-07-28T17:17:07.981-0700 INFO pool-1-thread-2 de.softwareforge.kafka.LoadCommand Loading table 'orders' into topic 'tpch.orders'...
2014-07-28T17:17:07.981-0700 INFO pool-1-thread-3 de.softwareforge.kafka.LoadCommand Loading table 'lineitem' into topic 'tpch.lineitem'...
2014-07-28T17:17:07.982-0700 INFO pool-1-thread-4 de.softwareforge.kafka.LoadCommand Loading table 'part' into topic 'tpch.part'...
2014-07-28T17:17:07.982-0700 INFO pool-1-thread-5 de.softwareforge.kafka.LoadCommand Loading table 'partsupp' into topic 'tpch.partsupp'...
2014-07-28T17:17:07.982-0700 INFO pool-1-thread-6 de.softwareforge.kafka.LoadCommand Loading table 'supplier' into topic 'tpch.supplier'...
2014-07-28T17:17:07.982-0700 INFO pool-1-thread-7 de.softwareforge.kafka.LoadCommand Loading table 'nation' into topic 'tpch.nation'...
2014-07-28T17:17:07.982-0700 INFO pool-1-thread-8 de.softwareforge.kafka.LoadCommand Loading table 'region' into topic 'tpch.region'...
2014-07-28T17:17:10.612-0700 ERROR pool-1-thread-8 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.region
2014-07-28T17:17:10.781-0700 INFO pool-1-thread-8 de.softwareforge.kafka.LoadCommand Generated 5 rows for table 'region'.
2014-07-28T17:17:10.797-0700 ERROR pool-1-thread-3 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.lineitem
2014-07-28T17:17:10.932-0700 ERROR pool-1-thread-1 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.customer
2014-07-28T17:17:11.068-0700 ERROR pool-1-thread-2 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.orders
2014-07-28T17:17:11.200-0700 ERROR pool-1-thread-6 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.supplier
2014-07-28T17:17:11.319-0700 INFO pool-1-thread-6 de.softwareforge.kafka.LoadCommand Generated 100 rows for table 'supplier'.
2014-07-28T17:17:11.333-0700 ERROR pool-1-thread-4 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.part
2014-07-28T17:17:11.466-0700 ERROR pool-1-thread-5 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.partsupp
2014-07-28T17:17:11.597-0700 ERROR pool-1-thread-7 kafka.producer.async.DefaultEventHandler Failed to collate messages by topic, partition due to: Failed to fetch topic metadata for topic: tpch.nation
2014-07-28T17:17:11.706-0700 INFO pool-1-thread-7 de.softwareforge.kafka.LoadCommand Generated 25 rows for table 'nation'.
2014-07-28T17:17:12.180-0700 INFO pool-1-thread-1 de.softwareforge.kafka.LoadCommand Generated 1500 rows for table 'customer'.
2014-07-28T17:17:12.251-0700 INFO pool-1-thread-4 de.softwareforge.kafka.LoadCommand Generated 2000 rows for table 'part'.
2014-07-28T17:17:12.905-0700 INFO pool-1-thread-2 de.softwareforge.kafka.LoadCommand Generated 15000 rows for table 'orders'.
2014-07-28T17:17:12.919-0700 INFO pool-1-thread-5 de.softwareforge.kafka.LoadCommand Generated 8000 rows for table 'partsupp'.
2014-07-28T17:17:13.877-0700 INFO pool-1-thread-3 de.softwareforge.kafka.LoadCommand Generated 60175 rows for table 'lineitem'.
Kafka now has a number of topics that are preloaded with data to query.
Step 3: Make the Kafka topics known to Presto
In your Presto installation, add a catalog properties file
~/.prestoadmin/catalog/kafka.properties for the Kafka connector.
This file lists the Kafka nodes and topics:
connector.name=kafka
kafka.nodes=localhost:9092
kafka.table-names=tpch.customer,tpch.orders,tpch.lineitem,tpch.part,tpch.partsupp,tpch.supplier,tpch.nation,tpch.region
kafka.hide-internal-columns=false
Now start Presto:
$ bin/launcher start
Because the Kafka tables all have the tpch. prefix in the configuration,
the tables are in the tpch schema. The connector is mounted into the
kafka catalog because the properties file is named kafka.properties.
Start the Presto CLI:
$ ./presto --catalog kafka --schema tpch
List the tables to verify that things are working:
presto:tpch> SHOW TABLES;
Table
----------
customer
lineitem
nation
orders
part
partsupp
region
supplier
(8 rows)
Step 4: Basic data querying
Kafka data is unstructured and it has no metadata to describe the format of the messages. Without further configuration, the Kafka connector can access the data and map it in raw form but there are no actual columns besides the built-in ones:
presto:tpch> DESCRIBE customer;
Column | Type | Extra | Comment
-------------------+---------+-------+---------------------------------------------
_partition_id | bigint | | Partition Id
_partition_offset | bigint | | Offset for the message within the partition
_segment_start | bigint | | Segment start offset
_segment_end | bigint | | Segment end offset
_segment_count | bigint | | Running message count per segment
_key | varchar | | Key text
_key_corrupt | boolean | | Key data is corrupt
_key_length | bigint | | Total number of key bytes
_message | varchar | | Message text
_message_corrupt | boolean | | Message data is corrupt
_message_length | bigint | | Total number of message bytes
(11 rows)
presto:tpch> SELECT count(*) FROM customer;
_col0
-------
1500
presto:tpch> SELECT _message FROM customer LIMIT 5;
_message
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
{"rowNumber":1,"customerKey":1,"name":"Customer#000000001","address":"IVhzIApeRb ot,c,E","nationKey":15,"phone":"25-989-741-2988","accountBalance":711.56,"marketSegment":"BUILDING","comment":"to the even, regular platelets. regular, ironic epitaphs nag e"}
{"rowNumber":3,"customerKey":3,"name":"Customer#000000003","address":"MG9kdTD2WBHm","nationKey":1,"phone":"11-719-748-3364","accountBalance":7498.12,"marketSegment":"AUTOMOBILE","comment":" deposits eat slyly ironic, even instructions. express foxes detect slyly. blithel
{"rowNumber":5,"customerKey":5,"name":"Customer#000000005","address":"KvpyuHCplrB84WgAiGV6sYpZq7Tj","nationKey":3,"phone":"13-750-942-6364","accountBalance":794.47,"marketSegment":"HOUSEHOLD","comment":"n accounts will have to unwind. foxes cajole accor"}
{"rowNumber":7,"customerKey":7,"name":"Customer#000000007","address":"TcGe5gaZNgVePxU5kRrvXBfkasDTea","nationKey":18,"phone":"28-190-982-9759","accountBalance":9561.95,"marketSegment":"AUTOMOBILE","comment":"ainst the ironic, express theodolites. express, even pinto bean
{"rowNumber":9,"customerKey":9,"name":"Customer#000000009","address":"xKiAFTjUsCuxfeleNqefumTrjS","nationKey":8,"phone":"18-338-906-3675","accountBalance":8324.07,"marketSegment":"FURNITURE","comment":"r theodolites according to the requests wake thinly excuses: pending
(5 rows)
presto:tpch> SELECT sum(cast(json_extract_scalar(_message, '$.accountBalance') AS double)) FROM customer LIMIT 10;
_col0
------------
6681865.59
(1 row)
The data from Kafka can be queried using Presto but it is not yet in
actual table shape. The raw data is available through the _message and
_key columns but it is not decoded into columns. As the sample data is
in JSON format, the JSON Functions and Operators built into Presto can be used
to slice the data.
Step 5: Add a topic description file
The Kafka connector supports topic description files to turn raw data into
table format. These files are located in the etc/kafka folder in the
Presto installation and must end with .json. It is recommended that
the file name matches the table name but this is not necessary.
Add the following file as etc/kafka/tpch.customer.json and restart Presto:
{
"tableName": "customer",
"schemaName": "tpch",
"topicName": "tpch.customer",
"key": {
"dataFormat": "raw",
"fields": [
{
"name": "kafka_key",
"dataFormat": "LONG",
"type": "BIGINT",
"hidden": "false"
}
]
}
}
The customer table now has an additional column: kafka_key.
presto:tpch> DESCRIBE customer;
Column | Type | Extra | Comment
-------------------+---------+-------+---------------------------------------------
kafka_key | bigint | |
_partition_id | bigint | | Partition Id
_partition_offset | bigint | | Offset for the message within the partition
_segment_start | bigint | | Segment start offset
_segment_end | bigint | | Segment end offset
_segment_count | bigint | | Running message count per segment
_key | varchar | | Key text
_key_corrupt | boolean | | Key data is corrupt
_key_length | bigint | | Total number of key bytes
_message | varchar | | Message text
_message_corrupt | boolean | | Message data is corrupt
_message_length | bigint | | Total number of message bytes
(12 rows)
presto:tpch> SELECT kafka_key FROM customer ORDER BY kafka_key LIMIT 10;
kafka_key
-----------
0
1
2
3
4
5
6
7
8
9
(10 rows)
The topic definition file maps the internal Kafka key (which is a raw long
in eight bytes) onto a Presto BIGINT column.
Step 6: Map all the values from the topic message onto columns
Update the etc/kafka/tpch.customer.json file to add fields for the
message and restart Presto. As the fields in the message are JSON, it uses
the json data format. This is an example where different data formats
are used for the key and the message.
{
"tableName": "customer",
"schemaName": "tpch",
"topicName": "tpch.customer",
"key": {
"dataFormat": "raw",
"fields": [
{
"name": "kafka_key",
"dataFormat": "LONG",
"type": "BIGINT",
"hidden": "false"
}
]
},
"message": {
"dataFormat": "json",
"fields": [
{
"name": "row_number",
"mapping": "rowNumber",
"type": "BIGINT"
},
{
"name": "customer_key",
"mapping": "customerKey",
"type": "BIGINT"
},
{
"name": "name",
"mapping": "name",
"type": "VARCHAR"
},
{
"name": "address",
"mapping": "address",
"type": "VARCHAR"
},
{
"name": "nation_key",
"mapping": "nationKey",
"type": "BIGINT"
},
{
"name": "phone",
"mapping": "phone",
"type": "VARCHAR"
},
{
"name": "account_balance",
"mapping": "accountBalance",
"type": "DOUBLE"
},
{
"name": "market_segment",
"mapping": "marketSegment",
"type": "VARCHAR"
},
{
"name": "comment",
"mapping": "comment",
"type": "VARCHAR"
}
]
}
}
Now for all the fields in the JSON of the message, columns are defined and
the sum query from earlier can operate on the account_balance column directly:
presto:tpch> DESCRIBE customer;
Column | Type | Extra | Comment
-------------------+---------+-------+---------------------------------------------
kafka_key | bigint | |
row_number | bigint | |
customer_key | bigint | |
name | varchar | |
address | varchar | |
nation_key | bigint | |
phone | varchar | |
account_balance | double | |
market_segment | varchar | |
comment | varchar | |
_partition_id | bigint | | Partition Id
_partition_offset | bigint | | Offset for the message within the partition
_segment_start | bigint | | Segment start offset
_segment_end | bigint | | Segment end offset
_segment_count | bigint | | Running message count per segment
_key | varchar | | Key text
_key_corrupt | boolean | | Key data is corrupt
_key_length | bigint | | Total number of key bytes
_message | varchar | | Message text
_message_corrupt | boolean | | Message data is corrupt
_message_length | bigint | | Total number of message bytes
(21 rows)
presto:tpch> SELECT * FROM customer LIMIT 5;
kafka_key | row_number | customer_key | name | address | nation_key | phone | account_balance | market_segment | comment
-----------+------------+--------------+--------------------+---------------------------------------+------------+-----------------+-----------------+----------------+---------------------------------------------------------------------------------------------------------
1 | 2 | 2 | Customer#000000002 | XSTf4,NCwDVaWNe6tEgvwfmRchLXak | 13 | 23-768-687-3665 | 121.65 | AUTOMOBILE | l accounts. blithely ironic theodolites integrate boldly: caref
3 | 4 | 4 | Customer#000000004 | XxVSJsLAGtn | 4 | 14-128-190-5944 | 2866.83 | MACHINERY | requests. final, regular ideas sleep final accou
5 | 6 | 6 | Customer#000000006 | sKZz0CsnMD7mp4Xd0YrBvx,LREYKUWAh yVn | 20 | 30-114-968-4951 | 7638.57 | AUTOMOBILE | tions. even deposits boost according to the slyly bold packages. final accounts cajole requests. furious
7 | 8 | 8 | Customer#000000008 | I0B10bB0AymmC, 0PrRYBCP1yGJ8xcBPmWhl5 | 17 | 27-147-574-9335 | 6819.74 | BUILDING | among the slyly regular theodolites kindle blithely courts. carefully even theodolites haggle slyly alon
9 | 10 | 10 | Customer#000000010 | 6LrEaV6KR6PLVcgl2ArL Q3rqzLzcT1 v2 | 5 | 15-741-346-9870 | 2753.54 | HOUSEHOLD | es regular deposits haggle. fur
(5 rows)
presto:tpch> SELECT sum(account_balance) FROM customer LIMIT 10;
_col0
------------
6681865.59
(1 row)
Now all the fields from the customer topic messages are available as
Presto table columns.
Step 7: Use live data
Presto can query live data in Kafka as it arrives. To simulate a live feed of data, this tutorial sets up a feed of live tweets into Kafka.
Setup a live Twitter feed
- Download the twistr tool
$ curl -o twistr https://repo1.maven.org/maven2/de/softwareforge/twistr_kafka_0811/1.2/twistr_kafka_0811-1.2.sh
$ chmod 755 twistr
- Create a developer account at https://dev.twitter.com/ and set up an access and consumer token.
- Create a
twistr.propertiesfile and put the access and consumer key and secrets into it:
twistr.access-token-key=...
twistr.access-token-secret=...
twistr.consumer-key=...
twistr.consumer-secret=...
twistr.kafka.brokers=localhost:9092
Create a tweets table on Presto
Add the tweets table to the ~/.prestoadmin/catalog/kafka.properties file:
connector.name=kafka
kafka.nodes=localhost:9092
kafka.table-names=tpch.customer,tpch.orders,tpch.lineitem,tpch.part,tpch.partsupp,tpch.supplier,tpch.nation,tpch.region,tweets
kafka.hide-internal-columns=false
Add a topic definition file for the Twitter feed as etc/kafka/tweets.json:
{
"tableName": "tweets",
"topicName": "twitter_feed",
"dataFormat": "json",
"key": {
"dataFormat": "raw",
"fields": [
{
"name": "kafka_key",
"dataFormat": "LONG",
"type": "BIGINT",
"hidden": "false"
}
]
},
"message": {
"dataFormat":"json",
"fields": [
{
"name": "text",
"mapping": "text",
"type": "VARCHAR"
},
{
"name": "user_name",
"mapping": "user/screen_name",
"type": "VARCHAR"
},
{
"name": "lang",
"mapping": "lang",
"type": "VARCHAR"
},
{
"name": "created_at",
"mapping": "created_at",
"type": "TIMESTAMP",
"dataFormat": "rfc2822"
},
{
"name": "favorite_count",
"mapping": "favorite_count",
"type": "BIGINT"
},
{
"name": "retweet_count",
"mapping": "retweet_count",
"type": "BIGINT"
},
{
"name": "favorited",
"mapping": "favorited",
"type": "BOOLEAN"
},
{
"name": "id",
"mapping": "id_str",
"type": "VARCHAR"
},
{
"name": "in_reply_to_screen_name",
"mapping": "in_reply_to_screen_name",
"type": "VARCHAR"
},
{
"name": "place_name",
"mapping": "place/full_name",
"type": "VARCHAR"
}
]
}
}
As this table does not have an explicit schema name, it will be placed
into the default schema.
Feed live data
Start the twistr tool:
$ java -Dness.config.location=file:$(pwd) -Dness.config=twistr -jar ./twistr
twistr connects to the Twitter API and feeds the “sample tweet” feed
into a Kafka topic called twitter_feed.
Now run queries against live data:
$ ./presto-cli --catalog kafka --schema default
presto:default> SELECT count(*) FROM tweets;
_col0
-------
4467
(1 row)
presto:default> SELECT count(*) FROM tweets;
_col0
-------
4517
(1 row)
presto:default> SELECT count(*) FROM tweets;
_col0
-------
4572
(1 row)
presto:default> SELECT kafka_key, user_name, lang, created_at FROM tweets LIMIT 10;
kafka_key | user_name | lang | created_at
--------------------+-----------------+------+-------------------------
494227746231685121 | burncaniff | en | 2014-07-29 14:07:31.000
494227746214535169 | gu8tn | ja | 2014-07-29 14:07:31.000
494227746219126785 | pequitamedicen | es | 2014-07-29 14:07:31.000
494227746201931777 | josnyS | ht | 2014-07-29 14:07:31.000
494227746219110401 | Cafe510 | en | 2014-07-29 14:07:31.000
494227746210332673 | Da_JuanAnd_Only | en | 2014-07-29 14:07:31.000
494227746193956865 | Smile_Kidrauhl6 | pt | 2014-07-29 14:07:31.000
494227750426017793 | CashforeverCD | en | 2014-07-29 14:07:32.000
494227750396653569 | FilmArsivimiz | tr | 2014-07-29 14:07:32.000
494227750388256769 | jmolas | es | 2014-07-29 14:07:32.000
(10 rows)
There is now a live feed into Kafka which can be queried using Presto.
Epilogue: Time stamps
The tweets feed that was set up in the last step contains a time stamp in
RFC 2822 format as created_at attribute in each tweet.
presto:default> SELECT DISTINCT json_extract_scalar(_message, '$.created_at')) AS raw_date
-> FROM tweets LIMIT 5;
raw_date
--------------------------------
Tue Jul 29 21:07:31 +0000 2014
Tue Jul 29 21:07:32 +0000 2014
Tue Jul 29 21:07:33 +0000 2014
Tue Jul 29 21:07:34 +0000 2014
Tue Jul 29 21:07:35 +0000 2014
(5 rows)
The topic definition file for the tweets table contains a mapping onto a
timestamp using the rfc2822 converter:
...
{
"name": "created_at",
"mapping": "created_at",
"type": "TIMESTAMP",
"dataFormat": "rfc2822"
},
...
This allows the raw data to be mapped onto a Presto timestamp column:
presto:default> SELECT created_at, raw_date FROM (
-> SELECT created_at, json_extract_scalar(_message, '$.created_at') AS raw_date
-> FROM tweets)
-> GROUP BY 1, 2 LIMIT 5;
created_at | raw_date
-------------------------+--------------------------------
2014-07-29 14:07:20.000 | Tue Jul 29 21:07:20 +0000 2014
2014-07-29 14:07:21.000 | Tue Jul 29 21:07:21 +0000 2014
2014-07-29 14:07:22.000 | Tue Jul 29 21:07:22 +0000 2014
2014-07-29 14:07:23.000 | Tue Jul 29 21:07:23 +0000 2014
2014-07-29 14:07:24.000 | Tue Jul 29 21:07:24 +0000 2014
(5 rows)
The Kafka connector contains converters for ISO 8601, RFC 2822 text formats and for number-based timestamps using seconds or miilliseconds since the epoch. There is also a generic, text-based formatter which uses Joda-Time format strings to parse text columns.