v1.10.7

• Java 8 upgrade
• Diagnostic health check endpoint
• Encryption
• Documentation update: encryption, kinesis producer, schema storage fundamentals, etc.

v1.10.6

• Binlog-connector upgrade
• Bug-fix: when using literal string for an option that accepts Regex, Regex characters are no longer special
• If master recovery is enabled, Maxwell cleans up old positions for the same server and client id

v1.10.5

• Shyko's binlog-connector is now the default and only replication
  backend available for maxwell.

v1.10.4

Notable changes:

• Shutdown hardening. If maxwell can't shut down (because the kafka
  producer is in a bad state and close() never terminates, for example),
  it would previously stall and process no messages. Now, shutdown is run
  in a separate thread and there is an additional watchdog thread which
  forcibly kills the maxwell process if it can't shut down within 10
  seconds.
• Initial support for running maxwell from java, rather then as its own
  process. This mode of operation is still experimental, but we'll
  accept PRs to improve it (thanks Geoff Lywood).
• Fix incorrect handling of negative (pre-epoch dates) when using
  binlog_connector mode (thanks Geoff Lywood).

v1.10.3

• tiny release to fix a units error in the replication.lag metric
  (subtracting seconds from milliseconds)

v1.10.2

• added metrics: "replication.queue.time" and "inflightmessages.count"
• renamed "time.overall" metric to "message.publish.time"
• documentation updates (thanks Chintan Tank)

v1.10.1

The observable changes in this minor release are a new configuration for Kafka/Kinesis producer to abort processing on publish errors, and support of Kafka 0.10.2. Also a bunch of good refactoring has been done for heartbeat processing. List of changes:

• Support Kafka 0.10.2
• Stop procesing RDS hearbeats
• Keep maxwell heartbeat going every 10 seconds when database is quiet
• Allow for empty double-quoted string literals for database schema changes
• Ignore Kafka/Kinesis producer errors based on new configuration ignore_producer_error

v1.10.0

This is a small release, primarily around a change to how schemas are
stored. Maxwell now stores the last_heartbeat_read with each entry
in the schemas table, making schema management more resilient to
cases where binlog numbers are reused, but means that you must take
care if you need to roll back to an earlier version. If you deploy
v1.10.0, then roll back to an earlier version, you should manually
update all schemas.last_heartbeat_read values to 0 before
redeploying v1.10.0 or higher.

Other minor changes:

• allow negative default numbers in columns
• only store final binlog position if it has changed
• blacklist internal aurora table `rds_heartbeat*'
• log4j version bump (allows for one entry per line JSON logging)

v1.10.0-rc1

v1.10.0-rc1

v1.9.0

Maxwell 1.9 adds one main feature: monitoring support, contributed by
Scott Ferguson. Multiple backends can be configured, read the updated
docs for full details.

There's also some bugfixes:

• filter DDL messages based on config
• determine newest schema from binlog order, not creation order
• add task manager to shutdown cleanly on error
• minor logging improvements