[BEAM-647] Fault-tolerant sideInputs via Broadcast variables

• Type: Bug
• Status: Resolved
• Priority: P2
• Resolution: Fixed
• Fix Version/s: 0.5.0
• Component/s: runner-spark
• Assignee: Kobi Salant
• Reporter: Amit Sela
• Created: 20/Sep/16 18:52
• Updated: 16/May/20 13:55
• Resolved: 10/Jan/17 17:29

Description

Following https://github.com/apache/incubator-beam/pull/909 which enables checkpointing to recover from failures, sideInputs (being implemented by broadcast variables) should be handled in a specific manner as described here: http://spark.apache.org/docs/1.6.2/streaming-programming-guide.html#accumulators-and-broadcast-variables.

This is a bit more complicated than Aggregators (via Accumulators) as they are implemented using a single "aggregating" Accumulator, while a pipeline may contain multiple sideInputs.

Issue Links

• links to: GitHub Pull Request #1624
• links to: GitHub Pull Request #1624