Load Data from Kafka

Overview

When running a CREATE PIPELINE ... KAFKA ... statement, you may need to make a secure connection to Kafka.

Use Secure Socket Layer (SSL) for the connection and Simple Authentication and Security Layer (SASL) to authenticate. Using SASL for authentication is optional.

• GSSAPI (Kerberos)

• PLAIN

• SCRAM-SHA-256

• SCRAM-SHA-512

• OAUTHBEARER SASL

Kerberos may used for authentication.

This topic assumes SSL and/or Kerberos have been set up, configured, and enabled on the Kafka brokers. For information on how to enable this functionality, see the SSL and SASL sections in the Kafka documentation.

Convert Java Keystore (JKS) to Privacy Enhanced Mail (PEM) Key

To use SSL encryption for SingleStore pipelines, JKS keys need to be converted to PEM keys.

1. Create the key and keystore. You will be prompted to enter details such as name, organizational unit, city, state, etc.

   Shell

   Copy

   keytool -genkey -keyalg RSA -keystore <keystore-name>.jks -storepass <password> -alias "<keystore-alias-name>"

   What is your first and last name?
    [Unknown]:- 
   What is the name of your organizational unit?
    [Unknown]:-
   What is the name of your organization?
    [Unknown]:-

2. Export the client certificate from the keystore using the same password as in step 1.

   Shell

   Copy

   keytool -exportcert -rfc -file <pem-name>.pem -alias <pem-name-alias> -keystore <keystore-name>.jks

   Enter keystore password:  <password>
   Certificate stored in file <pem-name>.pem

3. Import the client certificate to the truststore located on your Apache Server. Enter a new password for the keystore.

   Shell

   Copy

   keytool -keystore <truststore-name>.jks -alias <truststore-name-alias> -import -file <pem-name>.pem

   Enter keystore password:  
   Re-enter new password: 
   
   Trust this certificate? [no]:  yes
   Certificate was added to keystore

4. Convert the client keystore to Public-Key Cryptography Standards (PKCS12) format. The <key name>.jks and password are the same as in step 1.

   Shell

   Copy

   keytool -v -importkeystore -srckeystore <keystore-name>.jks -srcalias <keystore alias> -destkeystore <new-keystore-name>.p12 -deststoretype PKCS12

   Importing keystore <keyname>.jks to <new-keyname>.p12...
   Enter destination keystore password:  
   Re-enter new password: 
   Enter source keystore password:  
   [Storing new-keystore-name.p12]

5. Extract the client certificate key into a .pem file. Use the import password created in step 4.

   Shell

   Copy

   openssl pkcs12 -in <new-keyname>.p12 -nocerts -nodes > <keyname>.pem

   Enter Import Password:

Steps for Creating a Secure Connection

To create a secure connection from SingleStore to a Kafka cluster, follow these steps in order.

CREATE or REPLACE PIPELINE <pipeline_name>   
  AS LOAD DATA KAFKA "<Kafka cluster location>"   
  CONFIG '{"security.protocol":"SASL_SSL",  
         "sasl.mechanism":"OAUTHBEARER",
         "sasl.oauthbearer.client.id":"<CLIENT_ID>",    
         "sasl.oauthbearer.client.secret":<"CLIENT_SECRET>", 
         "sasl.oauthbearer.token.endpoint.url":"<ENDPOINT_URL>", 
         "sasl.oauthbearer.scope":"<SCOPE>"}'  
  INTO TABLE <table_name>;

SELECT * FROM information_schema.pipelines

Examples

The following examples make the following assumptions:

• Port 9092 is a plaintext endpoint

• Port 9093 is an SSL endpoint

• Port 9094 is a plaintext SASL endpoint

• Port 9095 is an SSL SASL endpoint

CREATE PIPELINE `kafka_plaintext`
AS LOAD DATA KAFKA 'host.example.com:9092/test'
CONFIG '{"security.protocol": "plaintext"}'
INTO table t;

CREATE PIPELINE `kafka_no_creds`
AS LOAD DATA KAFKA 'host.example.com:9092/test'
INTO table t;

CREATE PIPELINE `kafka_ssl`
AS LOAD DATA KAFKA 'host.example.com:9093/test'
CONFIG '{"security.protocol": "ssl",
"ssl.certificate.location": "/var/private/ssl/client_memsql_client.pem",
"ssl.key.location": "/var/private/ssl/client_memsql_client.key",
"ssl.ca.location": "/var/private/ssl/ca-cert.pem"}'
CREDENTIALS '{"ssl.key.password": "abcdefgh"}'
INTO table t;

CREATE PIPELINE `kafka_kerberos_no_ssl`
AS LOAD DATA KAFKA 'host.example.com:9094/test'
CONFIG '{""security.protocol": "sasl_plaintext",
"sasl.mechanism": "GSSAPI",
"sasl.kerberos.service.name": "kafka",
"sasl.kerberos.principal": "memsql/host.example.com@EXAMPLE.COM",
"sasl.kerberos.keytab": "/etc/krb5.keytab"}'
INTO table t

CREATE PIPELINE `kafka_kerberos_ssl`
AS LOAD DATA KAFKA 'host.example.com:9095/test'
CONFIG '{"security.protocol": "sasl_ssl",
"sasl.mechanism": "GSSAPI",
"ssl.certificate.location": "/var/private/ssl/client_memsql_client.pem",
"ssl.key.location": "/var/private/ssl/client_memsql_client.key",
"ssl.ca.location": "/var/private/ssl/ca-cert.pem",
"sasl.kerberos.service.name": "kafka",
"sasl.kerberos.principal": "memsql/host.example.com@EXAMPLE.COM",
"sasl.kerberos.keytab": "/etc/krb5.keytab"}'
CREDENTIALS '{"ssl.key.password": "abcdefgh"}'
INTO table t

CREATE PIPELINE `kafka_sasl_ssl_plain`
AS LOAD DATA KAFKA 'host.example.com:9095/test'
CONFIG '{"security.protocol": "sasl_ssl",
"sasl.mechanism": "PLAIN",
"ssl.certificate.location": "/var/private/ssl/client_memsql_client.pem",
"ssl.key.location": "/var/private/ssl/client_memsql_client.key",
"ssl.ca.location": "/var/private/ssl/ca-cert.pem",
"sasl.username": "kafka"}'
CREDENTIALS '{"ssl.key.password": "abcdefgh", "sasl.password": "metamorphosis"}'
INTO table t;

CREATE PIPELINE `kafka_sasl_plaintext_plain`
AS LOAD DATA KAFKA 'host.example.com:9094/test'
CONFIG '{"security.protocol": "sasl_plaintext",
"sasl.mechanism": "PLAIN",
"sasl.username": "kafka"}'
CREDENTIALS '{"sasl.password": "metamorphosis"}'
INTO table t;

CREATE PIPELINE `kafka_sasl_ssl_scram`
AS LOAD DATA KAFKA 'host.example.com:9095/test'
CONFIG '{"security.protocol": "sasl_ssl",
"sasl.mechanism": "SCRAM-SHA-512",
"ssl.certificate.location": "/var/private/ssl/client_memsql_client.pem",
"ssl.key.location": "/var/private/ssl/client_memsql_client.key",
"ssl.ca.location": "/var/private/ssl/ca-cert.pem",
"sasl.username": "kafka"}'
CREDENTIALS '{"ssl.key.password": "abcdefgh", "sasl.password": "metamorphosis"}'
INTO table t;

CREATE PIPELINE `kafka_sasl_plaintext_plain`
AS LOAD DATA KAFKA 'host.example.com:9094/test'
CONFIG '{"security.protocol": "sasl_plaintext",
"sasl.mechanism": "SCRAM-SHA-512",
"sasl.username": "kafka"}'
CREDENTIALS '{"sasl.password": "metamorphosis"}'
INTO table t;

CREATE or REPLACE PIPELINE <pipeline_name>   
AS LOAD DATA KAFKA "<Kafka cluster location>"  
CONFIG '{"security.protocol":"SASL_SSL",  
         "sasl.mechanism":"OAUTHBEARER",
         "sasl.oauthbearer.client.id":"<CLIENT_ID>",    
         "sasl.oauthbearer.client.secret":<"CLIENT_SECRET>", 
         "sasl.oauthbearer.token.endpoint.url":"<ENDPOINT_URL>", 
         "sasl.oauthbearer.scope":"<SCOPE>"}'  
INTO TABLE <table_name>;

Prerequisites

To complete this Quickstart, your environment must meet the following prerequisites:

• A working Kafka queue.

• SingleStore installation –or– a SingleStore cluster: You will connect to the database or cluster and create a pipeline to pull data from Kafka queue.

Part 1: Sending Messages to Kafka

In Kafka, create a topic named test and enter the following messages:

the quick
brown fox
jumped over
the lazy dog

In Part 2, you will create a pipeline in SingleStore to ingest these messages.

Part 2: Creating a Kafka Pipeline in SingleStore

Now that your Kafka topic contains some messages, you can create a new pipeline and ingest the messages.

At the SingleStore prompt, execute the following statements:

CREATE DATABASE quickstart_kafka;
USE quickstart_kafka;
CREATE TABLE messages (id text);

These statements create a new table and database that will be used for the Kafka pipeline. Before you can create the pipeline, you will need the IP address of the Kafka cluster.

To create the pipeline, execute the following statement, replacing <kafka-cluster-ip> with your Kafka cluster's IP address:

CREATE PIPELINE quickstart_kafka AS LOAD DATA KAFKA '<kafka-cluster-ip>/test' INTO TABLE messages;

If you are connecting with SSL, SASL, or Kerberos, you will need to specify additional clauses in your CREATE PIPELINE statement. See Securely Connect to Kafka for more information.

The CREATE PIPELINE statement just mentioned creates a new Kafka pipeline named quickstart_kafka, which reads messages from the test Kafka topic and writes it into the messages table. For more information, see the CREATE PIPELINE topic. If the statement was successful, you can test your pipeline. While you can start a pipeline after creating it, it’s always best to test it using a small set of data:

TEST PIPELINE quickstart_kafka LIMIT 1;

If this test was successful and no errors are present, then you are ready to try ingesting data. The following command will run one batch and commit the data to the SingleStore table messages.

START PIPELINE quickstart_kafka FOREGROUND LIMIT 1 BATCHES;

To verify that the data exists in the messages table as expected, execute the following statement. If it was successful, you should see a non-empty result set.

SELECT * FROM messages;

+--------------+
| id           |
+--------------+
| the quick    |
| brown fox    |
| jumped over  |
| the lazy dog |
+--------------+

Now you are ready to start your pipeline as a background process. SingleStore will automatically ingest new messages as they are put into Kafka.

START PIPELINE quickstart_kafka;

Now that the pipeline is up and running, send a few more messages to the Kafka topic. Enter the following lines into your topic:

Lorem ipsum
dolor sit amet

In the SingleStore terminal window, run the SELECT * FROM messages; statement again. Now you will see the following output:

SELECT * FROM messages;

+----------------+
| id             |
+----------------+
| lorem ipsum    |
| dolor sit amet |
| the quick      |
| brown fox      |
| jumped over    |
| the lazy dog   |
+----------------+

Now that your pipeline is running, you can check the status and history of it at any time by querying the PIPELINES_BATCHES_SUMMARY table.

SELECT * FROM information_schema.PIPELINES_BATCHES_SUMMARY;

This system view will give one row for every recent batch the pipeline has run, as well as at-a-glance performance and volume metrics. This information is extremely valuable for monitoring and understanding your production pipeline system.

Note: When a backup is restored all pipelines in that database will revert to the state (offsets, etc.) they were in when the target backup was generated. If the data isn't in Kafka anymore you will start from the earliest offsets available in Kafka.

Running kcat

Minimal syntax for running kcat is as follows.

kcat <mode> -b <broker> -t <topic>

When you run kcat, you may need to supply additional parameters, such as SASL settings to connect to your Kafka cluster. Specify these parameters using the -X option. Examples of these parameters are shown in the code samples below, which demonstrate how to connect to Confluent Cloud.

Verify connectivity to your Kafka Cluster

Running kcat in any mode, where the command returns successfully, is confirmation that you can connect. Following are two examples of how you can verify connectivity.

kcat -b <Confluent Cloud broker URL> -L \
-X security.protocol=SASL_SSL \
-X sasl.mechanism=PLAIN \
-X ssl.ca.location=<CA certificate file path> \
-X sasl.username=<CLUSTER_API_KEY> \
-X sasl.password=<CLUSTER_API_SECRET>

Metadata for all topics (from broker ...)

20 brokers:
...

10 topics:
...

kcat -C -b <Confluent Cloud broker URL> -t <topic> -o -1 -e \
-X security.protocol=SASL_SSL \
-X sasl.mechanism=PLAIN \
-X ssl.ca.location=<CA certificate file path> \
-X sasl.username=<CLUSTER_API_KEY> \
-X sasl.password=<CLUSTER_API_SECRET>

Verify data in your Kafka Cluster

Two situations where you may want to verify data in your Kafka cluster are:

• Prior to using SingleStore to consume Kafka messages via Pipelines or the Confluent Kafka Connector.

• After using SingleStore to produce Kafka messages (using SELECT ... INTO KAFKA ...).

In both situations, you can run kcat in -C mode for a given topic.

The following kcat command connects to a Confluent Cloud cluster using -C mode. The command outputs all of the messages for the specified topic, also showing the offset and partition for each message. The output is piped to the file kcat_output.txt. The -e option indicates that kcat should exit once it has consumed the last message from the topic.

kcat -C -b <Confluent Cloud broker URL> -t <topic> -e \
-f 'Partition: %p  Offset: %o  Message: %s\n' \
-X security.protocol=SASL_SSL \
-X sasl.mechanism=PLAIN \
-X ssl.ca.location=<CA certificate file path> \
-X sasl.username=<CLUSTER_API_KEY> \
-X sasl.password=<CLUSTER_API_SECRET>
> kcat_output.txt

CREATE TABLE numbers(a INT, b INT, c INT);

INSERT INTO numbers(a,b,c) VALUES (10,20,30);
INSERT INTO numbers(a,b,c) VALUES (40,50,60);
INSERT INTO numbers(a,b,c) VALUES (70,80,90);

SELECT a + 10, b + 20, c + 30
FROM numbers INTO KAFKA '<Confluent Cloud broker URL>/numbers-topic'
CONFIG '{
        "security.protocol":"SASL_SSL",
        "sasl.mechanism":"PLAIN",
        "ssl.ca.location":"<CA certificate file path>",
        "sasl.username":"<CLUSTER_API_KEY>"
        }'
CREDENTIALS '{"sasl.password":"<CLUSTER_API_SECRET>"}'
FIELDS TERMINATED BY ',' ENCLOSED BY '"' ESCAPED BY "\t"
LINES TERMINATED BY '}' STARTING BY '{';

Partition: 1  Offset: 0  Message: {20,40,60}
Partition: 1  Offset: 1  Message: {50,70,90}
Partition: 1  Offset: 2  Message: {80,100,120}
...

Error 1933: Cannot get source metadata for pipeline.

ERROR 1933 ER_EXTRACTOR_EXTRACTOR_GET_LATEST_OFFSETS: Cannot get source metadata for pipeline. Could not fetch Kafka metadata; are the Kafka brokers reachable from the Master Aggregator? ssl.ca.location is missing. Kafka error Local: Broker transport failure.

Error 1933 means that there is a communication error between SingleStore and the Kafka broker.

The cause could be something as simple as the Kafka broker is offline. Some things to check are:

• Broker Status: Confirm Kafka broker is in an online state.

• Port number: check to see if the port number is open and not in use by any other connection endpoint.

• Hostname: check if there is a typo in the hostname.

• Firewall: verify the firewall's rules are configured to allow incoming and outgoing traffic on the correct port(s).

• SSL configs: ensure that the SSL configurations are not missing and are correct.

How to Analyze the Debugging Log

This section shows how to enable debugging (if not already enabled), how to analyze the output to determine the cause, and assist in solving the error.

1. Enable the engine variable: pipelines_extractor_debug_logging for verbose error logging.

   SQL

   Copy

   SET GLOBAL pipelines_extractor_debug_logging=ON;

   Query OK, 0 rows affected (0.02 sec)

2. Flush the extractor pools.

   SQL

   Copy

   FLUSH EXTRACTOR POOLS;

   Query OK, 0 rows affected (0.00 sec)

3. Run the CREATE PIPELINE statement and wait to see if the pipeline fails. Below is a Kafka broker unreachable sample error.

   SQL

   Copy

   CREATE PIPELINE my_pipeline as 
    LOAD DATA KAFKA 'ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com/singlestore' 
    INTO TABLE messages;

   ERROR 1933 (HY000): Cannot get source metadata for pipeline.  
     Could not fetch Kafka metadata; are the Kafka brokers reachable from the Master Aggregator?  
     ssl.ca.location is missing. Kafka error Local: Broker transport failure.

4. Run SHOW WARNINGS. The output provides information as to what is causing the error. As shown in this example output, there is no communication with the Kafka broker on ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092. This may be due to no usable broker is present or the broker is not running.

   Note that if the issue is with an existing pipeline that is failing, the verbose debugging errors will be logged in the information_schema.pipelines_errors table.

   SQL

   Copy

   SHOW WARNINGS;

   +---------+------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
   | Level   | Code | Message                                                                                                                                                                               |
   +---------+------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+
   | Warning | 1933 |
   --- TRUNCATED ---
    ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap: Selected for cluster connection: application metadata request (broker has 567 connection attempt(s))
   CONNECT [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: 127.0.0.1:9092/bootstrap: Received CONNECT op
   STATE [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: 127.0.0.1:9092/bootstrap: Broker changed state INIT -> TRY_CONNECT
   BROADCAST [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: Broadcasting state change
   
   ...
   
   "FAIL [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap: Connect to ipv4#192.xxx.x.xx:9092 failed: 
   Connection refused (after 55ms in state CONNECT) (_TRANSPORT): identical to last error: error log suppressed
   STATE [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap: Broker changed state CONNECT -> DOWN"
   "CONNECT [thrd:main]: Not selecting any broker for cluster connection: still suppressed for 36ms: refresh unavailable topics
   METADATA [thrd:main]: Hinted cache of 1/1 topic(s) being queried
   METADATA [thrd:main]: Skipping metadata refresh of 1 topic(s): refresh unavailable topics: no usable brokers
   CONNECT [thrd:main]: Not selecting any broker for cluster connection: still suppressed for 36ms: no cluster connection"
   "CONNECT [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap: broker in state TRY_CONNECT connecting
   STATE [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap: Broker changed state TRY_CONNECT -> CONNECT
   BROADCAST [thrd:ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/bootstrap]: Broadcasting state change"

5. To turn off pipelines_extractor_debug_logging:

   SQL

   Copy

   SET GLOBAL pipelines_extractor_debug_logging = OFF;

When the variable pipelines_extractor_debug_logging is OFF, the warning messages are concise but still helpful. The output does not show all the available information. When this variable is ON, there is more information. Often an internet search will yield a solution based on the information provided in the output.

Confirm Listeners in Kafka server.properties is Correct

If the error: Cannot get source metadata for pipeline. Could not fetch Kafka metadata keeps occurring on pipeline creation, confirm that the hostname is correct in the Kafka server.properties file.

For example, a pipeline was created using the following DDL statement:

CREATE PIPELINE kafka
AS LOAD DATA KAFKA 
'ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com/singlestore'
INTO TABLE messages;

This pipeline creation statement resulted in the following error:

ERROR 1933 ER_EXTRACTOR_EXTRACTOR_GET_LATEST_OFFSETS: Cannot get 
source metadata for pipeline. Could not fetch Kafka metadata; are the Kafka 
brokers reachable from the Master Aggregator? Ssl.ca.location is missing. 
Kafka error Local Broker transport failure

To rectify this error, the server.properties file will need to be edited to reflect the correct hostname and to check for any other missing or incorrect settings. The location of the server.properties file may differ based on the version of Linux and Kafka that are being used.

1. Specify the broker’s hostname and include the port number it uses.

   SQL

   Copy

   listeners=PLAINTEXT://ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092

2. Restart the broker:

   Shell

   Copy

   ubuntu@kafka:~/kafka/config$ sudo systemctl restart kafka

3. Confirm the broker is running:

   Shell

   Copy

   ubuntu@kafka:~/kafka/config$ sudo systemctl status kafka | grep Active
      Active: active (running) since Wed 2022-12-21 01:29:29 UTC; 48s ago

4. Try running the create pipeline statement again:

   SQL

   Copy

   CREATE PIPELINE kafka
   AS LOAD DATA KAFKA'ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com/singlestore'
   INTO TABLE messages;

5. Check if the pipeline is ingesting data from Kafka:

   SQL

   Copy

   STOP PIPELINE kafka;
   
   Test PIPELINE kafka LIMIT 1;

   singlestore> Test PIPELINE kafka limit 1;
   +----------------------+------------------------------------------------------------------------------------------------------+
   | id                   | tweet                                                                                                |                                                                                                                                                                                          
   +----------------------+------------------------------------------------------------------------------------------------------+ 
   | 1612846720953581568  | {"created_at":1673367598,"favorite_count":0,"id":1612846720953581568,"retweet_count":0,"text":"RT    |
   |                      | @YE0NSSI: never getting over mr. jazz and boracut https://t.co/sHjsOaQ21S","username":"angelyoongo"} |
   +----------------------+------------------------------------------------------------------------------------------------------+
   1 row in set (0.49 sec)

Failed to get watermark offsets from Kafka with error Local: Unknown partition

Troubleshooting an Error 1933 error when trying to create a pipeline:

singlestore> create pipeline debug as load data kafka 
ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092, 
ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9093/multibroker' 
into table messages;
ERROR 1933 (HY000): Cannot get source metadata for pipeline.  
Failed to get watermark offsets from Kafka with error Local: 
Unknown partition

ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092 
ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9093

ubuntu@kafka:~/kafka/scripts$ sudo systemctl status broker1 | grep Active

     Active: active (running) since Wed 2023-01-04 01:01:56 UTC; 14min ago

ubuntu@kafka:~/kafka/scripts$ sudo systemctl status broker2 | grep Active

     Active: failed (Result: exit-code) since Wed 2023-01-04 01:01:04 UTC; 15min ago

ubuntu@kafka:~/kafka/scripts$ sudo systemctl start broker2

ubuntu@kafka:~/kafka/scripts$ sudo systemctl status broker2 | grep Active

  Active: active (running) since Wed 2023-01-04 01:26:21 UTC; 1min 14s ago

singlestore> CREATE PIPELIN debug AS LOAD DATA kafka 
'ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092, 
ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9093/multibroker' 
INTO TABLE messages;Query OK, 0 rows affected (0.35 sec)

ERROR 1933: Cannot get source metadata for pipeline.  Zero partitions for topic.  Topic may not exist, or it may still be being created.

The error shown below is  likely caused by a problem with the topic or the topic is still recovering from a broker reboot.

singlestore> system date
Thu Jan  5 23:38:30 UTC 2023
singlestore> CREATE PIPELINE debug AS LOAD DATA kafka 
'ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/debug' 
INTO TABLE messages;

ERROR 1933 (HY000): Cannot get source metadata for pipeline.  
Zero partitions for topic.  Topic may not exist, or it may still 
be being created.

Check Kafka’s server.log to cross-reference the timestamp when the error occurred with the information present in the Kafka logging. The log should provide helpful information regarding the error. In the  example below, there are traces about topic recovery:

ubuntu@kafka:~/kafka/logs$ rg debug server.log | rg -v multibroker | rg -v test | rg 23:3
[2023-01-05 23:38:29,975] INFO [LogLoader partition=debug-1, dir=/tmp/kafka-logs-broker0] Loading producer state till offset 51457 with message format version 2 (kafka.log.Log$)
[2023-01-05 23:38:29,975] INFO [LogLoader partition=debug-1, dir=/tmp/kafka-logs-broker0] Reloading from producer snapshot and rebuilding producer state from offset 51457 (kafka.log.Log$)
[2023-01-05 23:38:29,998] INFO [ProducerStateManager partition=debug-1] Loading producer state from snapshot file 'SnapshotFile(/tmp/kafka-logs-broker0/debug-1/00000000000000051457.snapshot,51457)' (kafka.log.ProducerStateManager)
[2023-01-05 23:38:30,034] INFO [LogLoader partition=debug-1, dir=/tmp/kafka-logs-broker0] Producer state recovery took 59ms for snapshot load and 0ms for segment recovery from offset 51457 (kafka.log.Log$)
[2023-01-05 23:38:30,049] INFO Completed load of Log(dir=/tmp/kafka-logs-broker0/debug-1, topicId=FeqyeH2ZR_28aMF1x40Z-g, topic=debug, partition=1, highWatermark=0, lastStableOffset=0, logStartOffset=0, logEndOffset=51457) with 1 segments in 107ms (5/17 loaded in /tmp/kafka-logs-broker0) (kafka.log.LogManager)
[2023-01-05 23:38:30,300] INFO [LogLoader partition=debug-0, dir=/tmp/kafka-logs-broker0] Loading producer state till offset 48543 with message format version 2 (kafka.log.Log$)
[2023-01-05 23:38:30,300] INFO [LogLoader partition=debug-0, dir=/tmp/kafka-logs-broker0] Reloading from producer snapshot and rebuilding producer state from offset 48543 (kafka.log.Log$)
[2023-01-05 23:38:30,300] INFO [ProducerStateManager partition=debug-0] Loading producer state from snapshot file 'SnapshotFile(/tmp/kafka-logs-broker0/debug-0/00000000000000048543.snapshot,48543)' (kafka.log.ProducerStateManager)
[2023-01-05 23:38:30,301] INFO [LogLoader partition=debug-0, dir=/tmp/kafka-logs-broker0] Producer state recovery took 1ms for snapshot load and 0ms for segment recovery from offset 48543 (kafka.log.Log$)
[2023-01-05 23:38:30,315] INFO Completed load of Log(dir=/tmp/kafka-logs-broker0/debug-0, topicId=FeqyeH2ZR_28aMF1x40Z-g, topic=debug, partition=0, highWatermark=0, lastStableOffset=0, logStartOffset=0, logEndOffset=48543) with 1 segments in 39ms (12/17 loaded in /tmp/kafka-logs-broker0) (kafka.log.LogManager)
[2023-01-05 23:38:35,951] INFO [Partition debug-0 broker=0] Log loaded for partition debug-0 with initial high watermark 48543 (kafka.cluster.Partition)
[2023-01-05 23:38:36,414] INFO [Partition debug-1 broker=0] Log loaded for partition debug-1 with initial high watermark 51457 (kafka.cluster.Partition)

A recovery topic is usually solved by running the CREATE PIPELINE command again.

singlestore> CREATE PIPELINE debug AS LOAD DATA kafka 
'ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092/debug'
INTO TABLE messages;
Query OK, 0 rows affected (0.05 sec)

Cannot get source metadata for pipeline.  Failed to get watermark offsets from Kafka with error Broker: Unknown topic or partition

If an error such as, Cannot get source metadata for pipeline.  Failed to get watermark offsets from Kafka with error Broker: Unknown topic or partition, arises in the information_schema.pipelines_errors table, check the Kafka broker(s) server.log around the same time as the error.

*** 1. row ***
            DATABASE_NAME: kafka
           PIPELINE_NAME: debug
     ERROR_UNIX_TIMESTAMP: 1672960138.923225
               ERROR_TYPE: Error
               ERROR_CODE: 1933
            ERROR_MESSAGE: Cannot get source metadata for pipeline.  
Failed to get watermark offsets from Kafka with error Broker: Unknown 
topic or partition
               ERROR_KIND: Extract
                STD_ERROR: 2023-01-05 23:07:51.811 Batch starting 
with new consumer.  rd_kafka_version_str: 1.8.2-PRE6-4-gecdd4d-dirty
2023-01-05 23:08:58.869 Failed to get watermark offsets from Kafka
with error Broker: Unknown topic or partition_$_SERVER_ERROR: 
Failed to get watermark offsets from Kafka with error Broker: 
Unknown topic or partition

In the Kafka server.log there are recovery type traces around the time of the error in the information_schema.pipelines_errors table.

[2023-01-05 23:08:58,276] INFO Kafka version: 3.0.0 (org.apache.kafka.common.utils.AppInfoParser)
[2023-01-05 23:08:58,277] INFO Kafka commitId: 8cb0a5e9d3441962 (org.apache.kafka.common.utils.AppInfoParser)
[2023-01-05 23:08:58,277] INFO Kafka startTimeMs: 1672960138244 (org.apache.kafka.common.utils.AppInfoParser)
[2023-01-05 23:08:58,279] INFO [KafkaServer id=0] started (kafka.server.KafkaServer)
[2023-01-05 23:08:58,581] INFO [BrokerToControllerChannelManager broker=0 name=alterIsr]: 
     Recorded new controller, from now on will use broker ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092 
     (id: 0 rack: null) (kafka.server.BrokerToControllerRequestThread)
[2023-01-05 23:08:58,642] INFO [BrokerToControllerChannelManager broker=0 name=forwarding]: 
     Recorded new controller, from now on will use broker ec2-xx-xxx-xx-xx.us-west-2.compute.amazonaws.com:9092 
     (id: 0 rack: null) (kafka.server.BrokerToControllerRequestThread)
[2023-01-05 23:08:59,279] INFO [Partition debug-0 broker=0] Log loaded for partition debug-0 with initial high watermark 48543 (kafka.cluster.Partition)

Once the broker is online and fully recovered, confirm if the problematic Kafka pipeline has stop printing the unknown topic or partition errors to the information_schema.pipelines_errors table.