v1.0.5

What's Changed

Backports of:

• stdserver: Improve health check ergonomics #1056
• sequtil: Remove extraneous return statement from LeaseGroup #1054
• stage: Add a configuration object to the stage package. #1053
• internal/target/schemawatch: added in configuration wiring for the RefreshDelay option #1048
• stage: updated the MarkApplied so that table updates happen in batches #1033

Full Changelog: v1.0.4...v1.0.5

v1.0.4

What's Changed

Enhancements

• checkpoint: Use core changefeed for fast notifications by @bobvawter in #1026

Full Changelog: v1.0.3...v1.0.4

v1.0.3

What's Changed

• Bugfix to Oracle Database column introspection query to elide hidden system columns
• Improvement to --limitLookhead flag implementation when set to small values

Full Changelog: v1.0.2...v1.0.3

v1.0.2

What's Changed

Enhancements

• checkpoint: Add option to limit the resolving range by @bobvawter in #984
  • The --limitLookahead flag can be used to limit the number of checkpoints to be considered when computing the resolving range. Note that this may cause replication to stall completely if older mutations cannot be applied.

Bug Fixes

• sequencer: Improve visibility of long-running background tasks by @bobvawter in #982
  • This change ensures that tasks from previous iterations of the replication loop have terminated, for example, relatively long-running database queries. If the previous loop cannot be cleaned up within --taskGracePeriod, the Replicator process will exit.

Full Changelog: v1.0.1...v1.0.2

v1.0.1

This is a re-released version which bumps the Go runtime to 1.22.6 to address a crash in goroutine 0.

What's Changed

Bug Fixes

• Prevent table reader from jumping backwards by @bobvawter in #959
  • It is possible for a reader to re-emit rows in subsequent cursors if it has reached an idle state and the bounds variable is updated without advancing the minimum value. This will happen if a slow target database delays the resolved timestamp bound from advancing before a new resolved timestamp is accepted for processing. This change updates the behavior of the table reader such that it will not re-emit rows, regardless of how the bounds variable is driven.
• Use transient accumulator in tableReader by @bobvawter in #958
  • If the retry loop in nextCursor retries, the reader's accumulator won't be cleared between retries. There's no overwhelming reason to want to save the accumulator instance between cycles, since we wind up moving the mutations into a MultiBatch. This change scopes the accumulator lifecycle to a single call to nextCursor() and queryOnce().

Enhancements

• Logging ergonomic enhancements by @bobvawter in #944
  • This change adds a metrics counter to identify spikes in workload error rates. The dashboard is updated to graph errors and warnings. Errors that are often a side-effect of some other malfunction are labeled with a "noisy" property to allow them to be filtered when log-mining. Log messages will now include the source of the logging call when trace mode is enabled.

Full Changelog: v1.0.0...v1.0.1

v1.0.0

This is the inaugural release of Replicator, formerly known as cdc-sink. We would like to thank everyone who's helped us get to this point, especially our amazing customers who are building high-scale data ecosystems. We're looking forward to more sources, targets, and integrations over the coming years.

Contributors

• @BramGruneir made their first contribution in #1
• @chriscasano made their first contribution in #26
• @robert-s-lee made their first contribution in #33
• @dependabot made their first contribution in #121
• @sravotto made their first contribution in #116
• @DuskEagle made their first contribution in #283
• @sean- made their first contribution in #306
• @otan made their first contribution in #398
• @glennfawcett made their first contribution in #537

Full Changelog: https://github.com/cockroachdb/replicator/commits/v1.0.0