Update streaming jobs in Hazelcast without data loss or interruptions.
Allows long-running jobs to be upgraded without data loss or interruption
Job Upgrade makes use of Jet state snapshots to address a variety of requirements, including modifications in business logic, bug fixes, and configuration changes.
Easy to Implement
Job upgrades can be triggered via a Java API or from a command line.
State snapshots of current jobs are taken and saved.
New classes or changes are distributed to the Jet nodes.
New versions are started, and data is read from the saved snapshot.
No Disruption to Customers
Everything listed above happens in milliseconds, with essentially no latency.
Allows A/B Testing on New Features
Multiple versions of a Jet Job can run concurrently, writing results to separated sinks.
After the testing period, pick the better version, and shut down the other.
Features
Bug Fixes
Introduce relevant filtering or transforming steps to streaming data-centric applications.
Business Logic Changes
Change business rules or add new operators as the business logic driving your streaming applications evolves.
Configuration Changes
Change IP addresses for sources or sinks without disrupting streaming services.
