GitHub user shanthoosh opened a pull request:
https://github.com/apache/samza/pull/792
SAMZA-1638: Recreate SystemProducer on
KafkaCheckpointManager.writeCheckpoint failures.
Retry loop in the existing `KafkaCheckpointManager` implementation retries
using the same `SystemProducer` instance on exception and does not recreate it.
When some irrecoverable exceptions occur within the `SystemProducer`, all
the subsequent produce message invocations on the `SystemProducer` instance
will fail. This had made the entire retry loop on `KafkaCheckpointManager`
pointless.
This patch consists of the following changes:
1. This patch addresses the above problem by recreating the
`SystemProducer` instance on failure and adds a unit test to verify the
functionality.
2. Minor code cleanup in classes: `TestKafkaCheckpointManager` and
`KafkaCheckpointManager`.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/shanthoosh/samza kafka_checkpoint_manager_fix
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/samza/pull/792.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #792
----
commit 1f2fc878a4d17f70d014f6442c6b81bee0137dd7
Author: Shanthoosh Venkataraman <spvenkat@...>
Date: 2018-11-02T23:27:26Z
Kafka checkpoint manager fix.
----
---
