Uploaded image for project: 'Percona Toolkit'
  1. Percona Toolkit
  2. PT-1483

[DOC] pt-table-checksum prints misleading warning about ROW format risks


    • Type: Bug
    • Status: Open
    • Priority: Medium
    • Resolution: Unresolved
    • Affects Version/s: 3.0.6
    • Fix Version/s: None
    • Component/s: Code
    • Labels:


      When the slave has binlog_format=ROW, the tool refuses to work with the following message:

      $ pt-table-checksum --user=msandbox --password=msandbox --databases="test" --recursion-method=hosts h=,u=msandbox,p=msandbox,P=20795
      Checking if all tables can be checksummed ...
      Starting checksum ...
      Replica bm-support01 has binlog_format ROW which could cause pt-table-checksum to break replication.  Please read "Replicas using row-based replication" in the LIMITATIONS section of the tool's documentation.  If you understand the risks, specify --no-check-binlog-format to disable this check.


      I think this is misleading, as the fact that the replica has ROW format only means that checksumming won't work as expected for 2-nd+ level slaves, if chain replication is used. But there is IMHO no reason why this should break replication.

      I think the message should just warn about the checksum results in chain replication only and that's all.


      Also, the documentation paragraph mentioned in the warning, doesn't say anything about breaking replicaiton:

      pt-table-checksum requires statement-based replication, and it sets binlog_format=STATEMENT on the master, but due to a MySQL limitation replicas do not honor this change. Therefore, checksums will not replicate past any replicas using row-based replication that are masters for further replicas.

      The tool automatically checks the binlog_format on all servers. See --[no]check-binlog-format .

      (Bug 899415)

        Smart Checklist




              • Assignee:
                przemyslaw.malkowski@percona.com Przemyslaw Malkowski
              • Votes:
                0 Vote for this issue
                2 Start watching this issue


                • Created: