FAQ

The most frequently asked questions ?

Could we deployed FilePulse connector in distributed mode ?

Connect File Pulse must be running locally to the machine hosting files to be ingested. It is recommend to deploy your connector in distributed mode. Multiple Kafka Connect workers can be deployed on the same machine and participating in the same cluster. The configured input directory will be scanned by the JVM running the SourceConnector. Then, all detected files will be scheduled amongs the tasks spread on your local cluster.

Is FilePulse connector fault-tolerant ?

Connect File Pulse guarantees no data loss by leveraging Kafka Connect fault-tolerance capabilities. Each task keeps a trace of the file offset of the last record written into Kafka. In case of a restart, tasks will continue where they stopped before crash. Note, that some duplicates maybe written into Kafka.

Is FilePulse connector could be used in place of other solutions like Logstash ?

Connect File Pulse has some features which are similar to the ones provided by Logstash codecs/filters. Filters like GrokFilter are actually strongly inspired from Logstash. For example you can use it to parse non-structured data like application logs.

However, Connect File Pulse has not to be originally designed to collect dynamic application log files.

Is FilePulse connector support SASL/SSL authentication mechanisms and can be deployed on Confluent Cloud ?

Yes, FilePulse connector can be deployed on any Kafka Cluster. However, the connector currently requires the use of an internal topic to synchronize the Connector instance and the Tasks that process files. For doing this, the connector will create both a producer and consumer clients that you must configured when running the connector with a secured Kafka Cluster.

To override the default configuration for the internal consumer and producer clients, you can use one of the following override prefixes :

  • internal.kafka.reporter.consumer.<consumer_property>
  • internal.kafka.reporter.producer.<producer_property>

What are the differences between FilePulse connector and others Kafka connectors for streaming files ?

The following table shows a simple comparison between Connect File Pulse and other solutions : Connect Spooldir and Connect FileStreams

Connect FilePulseConnect SpooldirConnect FileStreams
Connector Typesourcesourcesource / sink
LicenseApache License 2.0Apache License 2.0Apache License 2.0
Available on Confluent HubYESYESYES
Docker imageYESNONO
Delivery semanticsAt-least-onceAt-least-onceAt-most-once
Usable in productionYESYESNO
Supported file formats(out-of-the box)Delimited, Binary, JSON, Avro, XML (limited)Delimited, JSONText file
Support recursive directory scanYESNONO
Support Archive and Compressed filesYES (GZIP, TAR, ZIP)NONO
Source partitionsConfigurable (filename, path, filename+hash)filenamefilename
Support for multi-tasksYESYESNO
Support for worker distributed modeYES (requires a shared volume)NONO
Support for streaming log filesYESNOYES
Support for transformationSingle Message Transforms
Processing Filters (Grok, Append, JSON, etc)
Single Message Transforms* Single Message Transforms
Support for tracking processing progress of filesYES (using an internal topic)NONO