Kafka data source
Upsolver reads events from the specified Kafka topic in your Kafka cluster.
While setting up the data source, Upsolver scans the last 30 minutes of data in the topic in order find 100 records (to preview) and auto detect the content format.
If there is no data in the last 30 minutes, you will need to select a content format.
When a data source is created, it is populated with the sample 100 records from the preview. Once data is captured from the topic, this preview data is replaced with actual data.
This means that if you do not select to read from the start, the data source will initially appear with 100 historical records from the past 30 minutes, and these will be replaced when data is ingested from the topic.
If you create an output based on this data source, the preview data is not included.
By default the data in a data source is kept in Upsolver forever. If you specify a retention period in Upsolver (e.g. 30 days), this only applies to the data source.
If you create an output based on this data source, it will have its own retention, and these do not need to match (e.g. the output may be configured to retain the data for a year).
If you wish to customize the data source by selecting:
a specific compute cluster
a target storage option
retention options
See: Advanced​
1. From the Data Sources page, click New.
2. Select Kafka.
3. Enter in the host:port for your Kafka cluster.
If there are multiple hosts, the format should be host1:port1, host2:port2, ...
4. Select the Kafka topic to read from.
5. Name this data source.
6. Click Continue. The S3 Bucket Integration window will appear.
7. Click Launch Integration to launch the AWS CloudFormation page in a new tab.
8. Check the I acknowledge statement and click Create Stack.
Once the stack is created with the required resources, the message in Upsolver disappears automatically.
You can now use your Kafka topic data source.
1. From the Data Sources page, click New.
2. Select Kafka and then click Advanced.
3. Name this data source.
4. Select your Kafka version.
5. Enter in the host:port for your Kafka cluster.
If there are multiple hosts, the format should be host1:port1, host2:port2, ...
6. Select the Kafka topic to read from.
Click Show Advanced Options to configure:
Kafka Consumer Properties
Shards
Execution Parallelism
Real Time Statistics
Use SSL
End Read At
Store Raw Data
Compression
Otherwise, skip to step 15.
7. (Optional) Add any additional properties necessary for the Kafka consumer. The format should look as follows:
bootstrap.servers = HOST:PORTsecurity.protocol = SASL_SSLsasl.jaas.config = org.apache.kafka.common.security.plain.PlainLoginModule required username = "API_KEY" password = "SECRET";ssl.endpoint.identification.algorithm = httpssasl.mechanism = PLAIN
where:
HOST:PORT= the same Kafka host(s) entered in step 5API_KEY= an API key for your Kafka clusterSECRET= the corresponding secret for your API key
8. Specify the number of shards or parallel readers from Kafka (the more shards, the quicker the data processing).
Each shard in Upsolver can read from one or more partitions in Kafka.
The number of shards in Upsolver must be less than or equal to the number of partitions in Kafka.
Contact Upsolver Professional Services if you want to configure this option. Typically you need one Upsolver shard per 10-20 MBps of data.
9. Contact Upsolver Professional Services if you wish to configure the execution parallelism. This determines how many files will be created in the data source per minute and must be set to a value that is less than the number of shards.
10. Check Real Time Statistics to calculate the data source statistics in real time directly from the input topic.
This is relevant to lookup tables, when answers are required very fast and in real time.
11. Check Use SSL if your connection requires SSL.
Contact Upsolver Support to ensure that your SSL certificate is supported.
12. (Optional) Under End Read At, select a time to stop reading from the topic. This is useful if you wish to stop processing a topic.
13. Check Store Raw Data to store an additional copy of the data in its original format without any transformations.
14. Select the compression in the topic from the dropdown options.
If you set the compression to Auto Detect, the data that is ingested can be in multiple formats.
15. Decide whether or not to read the topic from the start. This could be a day or a week, depending on the Kafka setup.
If you do not select Read From Start, the topic will be read from the current time onwards and there will be no historical data in the data source.
16. Select the content format. This is typically auto-detected, but you can manually select a format.
If necessary, configure the content format options. See: Data formats​
17. From the dropdown, select a compute cluster (or create a new one) to run the calculation on.
A 
may appear. This warning can be ignored for POCs (proof of concept).
For production environments, this indicates that if you run your task retroactively (Start Ingestion From), your compute cluster will process a burst of additional tasks, possibly causing delays in outputs and lookup tables running on this cluster.
To prevent this, go to the Clusters page to edit your cluster and set the Additional Processing Units For Replay to a number greater than 0.
18. Select a target storage connection (or create a new one) to store the data read from the topic (output storage).
19. If you choose to check Enabled, specify a retention period in Upsolver for the data in minutes, hours, or days.
The data will be deleted permanently after this period elapses. By default, the data is kept forever.
20. Click Continue. A preview of the data will appear.
21. For CSV, select a Header.
22. Click Continue again.
Parsed data samples will appear and you can review the Sample Size, Parsed Successfully, and # of Errors.
23. (Optional) If there are any errors, click Back to change the settings as required.
24. Click Create.
You can now use your Kafka data source.
