Uploaded image for project: 'Debezium'
  1. Debezium
  2. DBZ-7256

Incremental Snapshot Start Lag

XMLWordPrintable

    • Icon: Bug Bug
    • Resolution: Unresolved
    • Icon: Major Major
    • under-triaging
    • None
    • core-library
    • None
    • False
    • None
    • False

       

      What Debezium connector do you use and what version?

      https://repo1.maven.org/maven2/io/debezium/debezium-connector-mysql/2.4.0.Final/debezium-connector-mysql-2.4.0.Final-plugin.tar.gz

      Relevant configuration?

      connector.class    "io.debezium.connector.mysql.MySqlConnector"

      incremental.snapshot.allow.schema.changes    "true"

      incremental.snapshot.chunk.size    "500000" (we've tried all sorts of different values for this)

      signal.enabled.channels "kafka"

      read.only    "true"

       

      What is the captured database version and mode of depoyment?

      MySQL 5.7

      What behaviour do you expect?

      I expect that when a connector reads a signal for an incremental snapshot, and logs... `Requested 'INCREMENTAL' snapshot of data collections [...]`

      that soon after we would get the STARTED notification in the notifications topic, and the incremental snapshot would begin.

      What behaviour do you see?

      We are getting lag anywhere from 15 minutes (consistent) to 3 hours (more extreme) between the time the connector logs that it got the incremental snapshot request, and when the Incremental Snapshot STARTS. We have noticed that the time taken to actually perform the snapshot, between STARTED and COMPLETED, is very reasonable, but the time to START is scaling very poorly.

       

      Do you have the connector logs, ideally from start till finish?

      Here is the correspondence with the logs and comments from Mario and Jiri. https://groups.google.com/g/debezium/c/5hQUAYQrRI0/m/b__-ERT_AAAJ?utm_medium=email&utm_source=footer

      How to reproduce the issue using our tutorial deployment?

      It is difficult to reproduce in a test env, as this issue seems to be a Production level performance issue. We are taking incremental snapshots on connectors that replicate thousands of tables. (this replication is tried and true, and has been running smoothly in prod for years, its just the Incremental Snapshotting piece that is not scaling).

      Feature request or enhancement

      For feature requests or enhancements, provide this information, please:

      Which use case/requirement will be addressed by the proposed feature?

      Faster Start Time for Incremental Snapshotting

            rh-ee-mvitale Mario Fiore Vitale
            drew-vz Drew von Zweck
            Jiri Pechanec, Mario Fiore Vitale
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

              Created:
              Updated: