Rahul Kumar created PHOENIX-7752:
------------------------------------
Summary: Feature to sync table data using PhoenixSyncTable tool on
inconsistency detection
Key: PHOENIX-7752
URL: https://issues.apache.org/jira/browse/PHOENIX-7752
Project: Phoenix
Issue Type: Sub-task
Reporter: Rahul Kumar
Assignee: Rahul Kumar
Fix For: 5.3.0, 5.2.1, 5.2.0
When chunk validation fails and the tool is running in repair mode, it performs
a two-pointer reconciliation to synchronize data between the source and target
chunk. The tool iterates through the source and target chunk row-by-row; if a
row is missing on the target, it issues a PUT mutation, and if an extraneous
row exists on the target, it issues a DELETE mutation. For rows present in both
locations, the same two-pointer logic is applied at the cell level, where the
tool identifies missing or extra cells and applies the corresponding PUT or
DELETE mutations. This repair process follows a best-effort model to ensure
high availability and progress: if a specific cell mutation fails, the tool
continues to sync the remaining cells in that row, and if a row-level mutation
fails, it gracefully proceeds to the next row in the sequence.
Repairing a row may not be possible due to delete markers on the target
cluster. If so, the repair failure is logged in the output table. Phoenix
compaction needs to run on the target cluster to remove these delete markers.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
