Stream ingestion

Apache Pinot lets users consume data from streams and push it directly into the database, in a process known as stream ingestion. Stream Ingestion makes it possible to query data within seconds of publication.

Stream Ingestion provides support for checkpoints for preventing data loss.

Setting up Stream ingestion involves the following steps:

1. Create schema configuration

2. Create table configuration

3. Upload table and schema spec

Let's take a look at each of the steps in more detail.

Let us assume the data to be ingested is in the following format:

{"studentID":205,"firstName":"Natalie","lastName":"Jones","gender":"Female","subject":"Maths","score":3.8,"timestamp":1571900400000}
{"studentID":205,"firstName":"Natalie","lastName":"Jones","gender":"Female","subject":"History","score":3.5,"timestamp":1571900400000}
{"studentID":207,"firstName":"Bob","lastName":"Lewis","gender":"Male","subject":"Maths","score":3.2,"timestamp":1571900400000}
{"studentID":207,"firstName":"Bob","lastName":"Lewis","gender":"Male","subject":"Chemistry","score":3.6,"timestamp":1572418800000}
{"studentID":209,"firstName":"Jane","lastName":"Doe","gender":"Female","subject":"Geography","score":3.8,"timestamp":1572505200000}
{"studentID":209,"firstName":"Jane","lastName":"Doe","gender":"Female","subject":"English","score":3.5,"timestamp":1572505200000}
{"studentID":209,"firstName":"Jane","lastName":"Doe","gender":"Female","subject":"Maths","score":3.2,"timestamp":1572678000000}
{"studentID":209,"firstName":"Jane","lastName":"Doe","gender":"Female","subject":"Physics","score":3.6,"timestamp":1572678000000}
{"studentID":211,"firstName":"John","lastName":"Doe","gender":"Male","subject":"Maths","score":3.8,"timestamp":1572678000000}
{"studentID":211,"firstName":"John","lastName":"Doe","gender":"Male","subject":"English","score":3.5,"timestamp":1572678000000}
{"studentID":211,"firstName":"John","lastName":"Doe","gender":"Male","subject":"History","score":3.2,"timestamp":1572854400000}
{"studentID":212,"firstName":"Nick","lastName":"Young","gender":"Male","subject":"History","score":3.6,"timestamp":1572854400000}

Create Schema Configuration

Schema defines the fields along with their data types. The schema also defines whether fields serve as dimensions , metrics or timestamp. For more details on schema configuration, see creating a schema.

For our sample data, the schema configuration looks like this:

/tmp/pinot-quick-start/transcript-schema.json

{
  "schemaName": "transcript",
  "dimensionFieldSpecs": [
    {
      "name": "studentID",
      "dataType": "INT"
    },
    {
      "name": "firstName",
      "dataType": "STRING"
    },
    {
      "name": "lastName",
      "dataType": "STRING"
    },
    {
      "name": "gender",
      "dataType": "STRING"
    },
    {
      "name": "subject",
      "dataType": "STRING"
    }
  ],
  "metricFieldSpecs": [
    {
      "name": "score",
      "dataType": "FLOAT"
    }
  ],
  "dateTimeFieldSpecs": [{
    "name": "timestamp",
    "dataType": "LONG",
    "format" : "1:MILLISECONDS:EPOCH",
    "granularity": "1:MILLISECONDS"
  }]
}

Create Table Configuration

The next step is to create a table where all the ingested data will flow and can be queried. Unlike batch ingestion, table configuration for real-time ingestion also triggers the data ingestion job. For a more detailed overview of tables, see the table reference.

The real-time table configuration consists of the following fields:

• tableName - The name of the table where the data should flow

• tableType - The internal type for the table. Should always be set to REALTIME for realtime ingestion

• segmentsConfig -

• tableIndexConfig - defines which column to use for indexing along with the type of index. For full configuration, see [Indexing Configs]. It has the following required fields -

  • loadMode - specifies how the segments should be loaded. Should beheap or mmap. Here's the difference between both the configs

    • mmap: Segments are loaded onto memory-mapped files. This is the default mode.

    • heap: Segments are loaded into direct memory. Note, 'heap' here is a legacy misnomer, and it does not imply JVM heap. This mode should only be used when we want faster performance than memory-mapped files, and are also sure that we will never run into OOM.

  • streamConfig - specifies the data source along with the necessary configs to start consuming the real-time data. The streamConfig can be thought of as the equivalent to the job spec for batch ingestion. The following options are supported:

Config key	Description	Supported values
streamType	The streaming platform from which to consume the data	kafka
stream.[streamType].consumer.type	Whether to use per partition low-level consumer or high-level stream consumer	lowLevel - Consume data from each partition with offset management highLevel - Consume data without control over the partitions
stream.[streamType].topic.name	The datasource (e.g. topic, data stream) from which to consume the data	String
stream.[streamType].decoder.class.name	Name of the class to be used for parsing the data. The class should implement org.apache.pinot.spi.stream.StreamMessageDecoder interface	String. Available options: org.apache.pinot.plugin.inputformat.json.JSONMessageDecoder org.apache.pinot.plugin.inputformat.avro.KafkaAvroMessageDecoder org.apache.pinot.plugin.inputformat.avro.SimpleAvroMessageDecoder org.apache.pinot.plugin.inputformat.avro.confluent.KafkaConfluentSchemaRegistryAvroMessageDecoder org.apache.pinot.plugin.inputformat.csv.CSVMessageDecoder org.apache.pinot.plugin.inputformat.protobuf.ProtoBufMessageDecoder
stream.[streamType].consumer.factory.class.name	Name of the factory class to be used to provide the appropriate implementation of low level and high level consumer as well as the metadata	String. Available options: org.apache.pinot.plugin.stream.kafka09.KafkaConsumerFactory org.apache.pinot.plugin.stream.kafka20.KafkaConsumerFactory org.apache.pinot.plugin.stream.kinesis.KinesisConsumerFactory org.apache.pinot.plugin.stream.pulsar.PulsarConsumerFactory
stream.[streamType].consumer.prop.auto.offset.reset	Determines the offset from which to start the ingestion	smallest largest or timestamp in milliseconds
topic.consumption.rate.limit	Determines the upper bound for consumption rate for the whole topic. Having a consumption rate limiter is beneficial in case the stream message rate has a bursty pattern which leads to long GC pauses on the Pinot servers. The rate limiter can also be considered as a safeguard against excessive ingestion of realtime tables.	Double. The values should be greater than zero.

The following flush threshold settings are also supported:

Config key	Description	Supported values
realtime.segment.flush.threshold.time	Time threshold that will keep the realtime segment open for before we complete the segment. Noted that this time should be smaller than the Kafka retention period configured for the corresponding topic.
realtime.segment.flush.threshold.rows	Row count flush threshold for realtime segments. This behaves in a similar way for HLC and LLC. For HLC, since there is only one consumer per server, this size is used as the size of the consumption buffer and determines after how many rows we flush to disk. For example, if this threshold is set to two million rows, then a high level consumer would have a buffer size of two million. If this value is set to 0, then the consumers adjust the number of rows consumed by a partition such that the size of the completed segment is the desired size (unless threshold.time is reached first)
realtime.segment.flush.threshold.segment.size	The desired size of a completed realtime segment. This config is used only if realtime.segment.flush.threshold.rows is set to 0.

You can also specify additional configs for the consumer directly into the streamConfigs.

For our sample data and schema, the table config will look like this:

{
  "tableName": "transcript",
  "tableType": "REALTIME",
  "segmentsConfig": {
    "timeColumnName": "timestamp",
    "timeType": "MILLISECONDS",
    "schemaName": "transcript",
    "replicasPerPartition": "1"
  },
  "tenants": {},
  "tableIndexConfig": {
    "loadMode": "MMAP",
    "streamConfigs": {
      "streamType": "kafka",
      "stream.kafka.consumer.type": "lowlevel",
      "stream.kafka.topic.name": "transcript-topic",
      "stream.kafka.decoder.class.name": "org.apache.pinot.plugin.stream.kafka.KafkaJSONMessageDecoder",
      "stream.kafka.consumer.factory.class.name": "org.apache.pinot.plugin.stream.kafka20.KafkaConsumerFactory",
      "stream.kafka.broker.list": "localhost:9876",
      "realtime.segment.flush.threshold.time": "3600000",
      "realtime.segment.flush.threshold.rows": "50000",
      "stream.kafka.consumer.prop.auto.offset.reset": "smallest"
    }
  },
  "metadata": {
    "customConfigs": {}
  }
}

Upload schema and table config

Now that we have our table and schema configurations, let's upload them to the Pinot cluster. As soon as the configs are uploaded, pinot will start ingesting available records from the topic.

Docker

docker run \
    --network=pinot-demo \
    -v /tmp/pinot-quick-start:/tmp/pinot-quick-start \
    --name pinot-streaming-table-creation \
    apachepinot/pinot:latest AddTable \
    -schemaFile /tmp/pinot-quick-start/transcript-schema.json \
    -tableConfigFile /tmp/pinot-quick-start/transcript-table-realtime.json \
    -controllerHost pinot-quickstart \
    -controllerPort 9000 \
    -exec

Launcher Script

bin/pinot-admin.sh AddTable \
    -schemaFile /path/to/transcript-schema.json \
    -tableConfigFile /path/to/transcript-table-realtime.json \
    -exec

Tuning Stream Config

Throttling Stream Consumption

There are some scenarios where the message rate in the input stream has a bursty nature which can lead to long GC pauses on the Pinot servers or affect the ingestion rate of other realtime tables on the same server. In such scenarios, you should throttle the consumption rate during stream ingestion.

Stream consumption throttling can be tuned using the stream config topic.consumption.rate.limit which indicates the upper bound on the message rate for the entire topic.

Here is the sample configuration on how to configure the consumption throttling:

{
  "tableName": "transcript",
  "tableType": "REALTIME",
  ...
  "tableIndexConfig": {
    "loadMode": "MMAP",
    "streamConfigs": {
      "streamType": "kafka",
      "stream.kafka.consumer.type": "lowlevel",
      "stream.kafka.topic.name": "transcript-topic",
      ...
      "topic.consumption.rate.limit": 1000
    }
  },
  ...
}

Some things to keep in mind while tuning this config are:

1. Since this config applied to the entire topic, internally, this rate is divided by the number of partitions in the topic and applied to each partition's consumer.

2. In case of multi-tenant deployment (where you have more than 1 table in the same server instance), you need to make sure that the rate limit on one table doesn't step on/starve the rate limiting of another table. So, when there is more than 1 table on the same server (which is most likely to happen), you may need to re-tune the throttling threshold for all the streaming tables.

Once throttling is enabled for a table, you can verify by searching for a log that looks similar to:

A consumption rate limiter is set up for topic <topic_name> in table <tableName> with rate limit: <rate_limit> (topic rate limit: <topic_rate_limit>, partition count: <partition_count>)

In addition, you can monitor the consumption rate utilization with the metric COSUMPTION_QUOTA_UTILIZATION.

Note that any configuration change for topic.consumption.rate.limit in the stream config will NOT take effect immediately. The new configuration will be picked up from the next consuming segment. In order to enforce the new configuration, you need to trigger forceCommit APIs. Please refer to Pause Stream Ingestion for more details.

$ curl -X POST {controllerHost}/tables/{tableName}/forceCommit

Custom Ingestion Support

We are working on support for other ingestion platforms, but you can also write your own ingestion plugin if it is not supported out of the box. For a walkthrough, see Stream Ingestion Plugin.

Pause Stream Ingestion

There are some scenarios in which you may want to pause the realtime ingestion while your table is available for queries. For example if there is a problem with the stream ingestion, while you are troubleshooting the issue, you still want the queries to be executed on the already ingested data. For these scenarios, you can first issue a Pause request to a Controller host. After troubleshooting with the stream is done, you can issue another request to Controller to resume the consumption.

$ curl -X POST {controllerHost}/tables/{tableName}/pauseConsumption
$ curl -X POST {controllerHost}/tables/{tableName}/resumeConsumption

When a Pause request is issued, Controller instructs the realtime servers hosting your table to commit their consuming segments immediately. However, the commit process may take some time to complete. Please note that Pause and Resume requests are async. OK response means that instructions for pausing or resuming has been successfully sent to the realtime server. If you want to know if the consumptions actually stopped or resumed, you can issue a pause status request.

$ curl -X POST {controllerHost}/tables/{tableName}/pauseStatus

It's worth noting that consuming segments on realtime servers are stored in volatile memory, and their resources are allocated when the consuming segments are first created. These resources cannot be altered if consumption parameters are changed midway through consumption. It may therefore take hours before these changes take effect. Furthermore, if the parameters are changed in an incompatible way (for example, changing the underlying stream with a completely new set of offsets, or changing the stream endpoint from which to consume messages, etc.), it will result in the table getting into an error state.

Pause and resume feature comes to the rescue here. When a Pause request is issued by the operator, consuming segments are committed without starting new mutables ones. Instead, new mutable segments are started only when the Resume request is issued. This mechanism provides the operators as well as developers with more flexibility. It also enables Pinot to be more resilient to the operational and functional constraints imposed by underlying streams.

There is another feature called "Force Commit" which utilizes the primitives of pause and resume feature. When the operator issues a force commit request, the current mutable segments will be committed and new ones started right away. Operators can now use this feature for all compatible table config parameter changes to take effect immediately.

$ curl -X POST {controllerHost}/tables/{tableName}/forceCommit

(v 0.12.0+) Once submitted, the forceCommit API returns a jobId that can be used to get the current progress of the forceCommit operation. A sample response and status API call:

$ curl -X POST {controllerHost}/tables/{tableName}/forceCommit
{
  "forceCommitJobId": "6757284f-b75b-45ce-91d8-a277bdbc06ae",
  "forceCommitStatus": "SUCCESS",
  "jobMetaZKWriteStatus": "SUCCESS"
}

$ curl -X GET {controllerHost}/tables/forceCommitStatus/6757284f-b75b-45ce-91d8-a277bdbc06ae
{
  "jobId": "6757284f-b75b-45ce-91d8-a277bdbc06ae",
  "segmentsForceCommitted": "[\"airlineStats__0__0__20230119T0700Z\",\"airlineStats__1__0__20230119T0700Z\",\"airlineStats__2__0__20230119T0700Z\"]",
  "submissionTimeMs": "1674111682977",
  "numberOfSegmentsYetToBeCommitted": 0,
  "jobType": "FORCE_COMMIT",
  "segmentsYetToBeCommitted": [],
  "tableName": "airlineStats_REALTIME"
}

For incompatible parameter changes, an option is added to the resume request to handle the case of a completely new set of offsets. Operators can now follow a three-step process: First, issue a Pause request. Second, change the consumption parameters. Finally, issue the Resume request with the appropriate option. These steps will preserve the old data and allow the new data to be consumed immediately. All through the operation, queries will continue to be served.

$ curl -X POST {controllerHost}/tables/{tableName}/resumeConsumption?resumeFrom=smallest
$ curl -X POST {controllerHost}/tables/{tableName}/resumeConsumption?resumeFrom=largest

Handling partition changes in Streams

If a Pinot table is configured to consume using a Low Level (partition-based) stream type, then it is possible that the partitions of the table change over time. In Kafka, for example, the number of partitions may increase. In Kinesis, the number of partitions may increase or decrease -- some partitions could be merged to create a new one, or existing partitions split to create new ones.

Pinot runs a periodic task called RealtimeSegmentValidationManager that monitors such changes and starts consumption on new partitions (or stops consumptions from old ones) as necessary. Since this is a periodic task that is run on the controller, it may take some time for Pinot to recognize new partitions and start consuming from them. This may delay the data in new partitions appearing in the results that pinot returns.

If it is desired to recognize the new partitions sooner, then you can manually trigger the periodic task so as to recognize such data immediately.

Inferring Ingestion Status of Realtime Tables

Often, it is important to understand the rate of ingestion of data into your realtime table. This is commonly done by looking at the consumption "lag" of the consumer. The lag itself can be observed in many dimensions. Pinot supports observing consumption lag along the offset dimension and time dimension, whenever applicable (as it depends on the specifics of the connector).

The ingestion status of a connector can be observed by querying either the /consumingSegmentsInfo API or the table's /debug API, as shown below:

# GET /tables/{tableName}/consumingSegmentsInfo
curl -X GET "http://<controller_url:controller_admin_port>/tables/meetupRsvp/consumingSegmentsInfo" -H "accept: application/json"

# GET /debug/tables/{tableName}
curl -X GET "http://localhost:9000/debug/tables/meetupRsvp?type=REALTIME&verbosity=1" -H "accept: application/json"

A sample response from a Kafka based realtime table is shown below. The ingestion status is displayed for each of the CONSUMING segments in the table.

{
  "_segmentToConsumingInfoMap": {
    "meetupRsvp__0__0__20221019T0639Z": [
      {
        "serverName": "Server_192.168.0.103_7000",
        "consumerState": "CONSUMING",
        "lastConsumedTimestamp": 1666161593904,
        "partitionToOffsetMap": { // <<-- Deprecated. See currentOffsetsMap for same info
          "0": "6"
        },
        "partitionOffsetInfo": {
          "currentOffsetsMap": {
            "0": "6" // <-- Current consumer position
          },
          "latestUpstreamOffsetMap": {
            "0": "6"  // <-- Upstream latest position
          },
          "recordsLagMap": {
            "0": "0"  // <-- Lag, in terms of #records behind latest
          },
          "recordsAvailabilityLagMap": {
            "0": "2"  // <-- Lag, in terms of time
          }
        }
      }
    ],

Term	Description
currentOffsetsMap	Current consuming offset position per partition
latestUpstreamOffsetMap	(Wherever applicable) Latest offset found in the upstream topic partition
recordsLagMap	(Whenever applicable) Defines how far behind the current record's offset / pointer is from upstream latest record. This is calculated as the difference between the latestUpstreamOffset and currentOffset for the partition when the lag computation request is made.
recordsAvailabilityLagMap	(Whenever applicable) Defines how soon after record ingestion was the record consumed by Pinot. This is calculated as the difference between the time the record was consumed and the time at which the record was ingested upstream.

Monitoring Realtime Ingestion

Realtime ingestion includes 3 stages of message processing: Decode, Transform and Index.

In each of these stages, a failure can happen which may or may not result in an ingestion failure. The following metrics are available to investigation ingestion issues:

1. Decode stage -> error here is recorded as INVALID_REALTIME_ROWS_DROPPED

2. Transform stage -> possible errors here are:

   1. When a message gets dropped due to the FILTER transform, it is recorded as REALTIME_ROWS_FILTERED

   2. When the transform pipeline sets the $INCOMPLETE_RECORD_KEY$ key in the message, it is recorded as INCOMPLETE_REALTIME_ROWS_CONSUMED , only when continueOnError configuration is enabled. If the continueOnError is not enabled, the ingestion fails.

3. Index stage -> When there is failure at this stage, the ingestion typically stops and marks the partition as ERROR.

There is yet another metric called ROWS_WITH_ERROR which is the sum of all error counts in the 3 stages above.

Furthermore, the metric REALTIME_CONSUMPTION_EXCEPTIONS gets incremented whenever there is a transient/permanent stream exception seen during consumption.

These metrics can be used to understand why ingestion failed for a particular table partition before diving into the server logs.