SEZ9 commented on issue #13223:
URL: https://github.com/apache/seatunnel/issues/13223#issuecomment-6870090188

   Thanks for the detailed follow-up, @SEPURI-SAI-KRISHNA. Taking the three 
rows in order:
   
   **Iceberg.** The `RANDOM` reasoning for the append writers holds: 
`UnpartitionedWriter` and `PartitionedAppendWriter` append every record as a 
data row whatever its `RowKind`, so there is no delete path and no ordering 
between changes to preserve, and your 3500-row result at replica 1 and 4 
confirms it. `KEYED` on the first key column for the delta writers is fine too. 
Two asks: keep the replica 1 vs 4 append measurement in the PR, as a test or at 
least in the description with the input shape, and make sure the 
repeated-changes case at replica 4 asserts the correct count with the fix in 
place, not only that random routing yields 139 instead of 50. Waiting for Phase 
1 and rebasing on the delta-writer change if it lands first is the right order.
   
   **JDBC XA.** Fine to keep it `UNSUPPORTED` until the two PRs you named land, 
with the same-layout restore at replica 2 in that PR. Nothing further needed 
now.
   
   **JDBC UNIQUE gap.** Routing by the `primary_keys` value in every case 
(configured, derived from the primary key, or derived from the UNIQUE key) is 
the right fix, and the test with configured `primary_keys` differing from the 
table primary key covers the important case. Please coordinate the rebase with 
@goutamadwant so your PR sits cleanly on top of the `byte[]` hash and 
key-refresh changes without duplicating them. For the negative case, relying on 
`testKeyedWriterNeedsAValidKey` and `testOneReplicaAcceptsEveryModeWithoutAKey` 
in Phase 1 is fine; just include the replica 1, 2 and 4 runs in the JDBC E2E as 
planned.
   
   I will update the table in the issue body as each PR opens, so posting them 
here as you do now is enough.
   
   <!-- streview-comment:2328 -->


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to