Caideyipi commented on PR #18623:
URL: https://github.com/apache/iotdb/pull/18623#issuecomment-6621778996

   Completed one independent re-review of the full current diff and relevant 
context at `f7d21bae403eed07f9ccdb5f8b0f242b11f5e2a1`, against 
`master/e6b44b57e49e1e7f0ecf5d6c62fa7e1f69a9a869`.
   
   - **[P1] Rebuild snapshot staging before resuming a request checkpoint.** 
[PipeLogicalBackupTool.java:1025-1046 at the reviewed 
revision](https://github.com/apache/iotdb/blob/f7d21bae403eed07f9ccdb5f8b0f242b11f5e2a1/iotdb-client/cli/src/main/java/org/apache/iotdb/tool/pipe/PipeLogicalBackupTool.java#L1025-L1046)
 skipped already acknowledged file pieces after opening a fresh receiver 
session. That session has discarded the previous temporary files. Resuming a 
schema/config snapshot therefore returned offset-reset status 1803 for a later 
piece, or failed sealing because files were missing; repeated `--resume` 
attempts could not advance. The resume path must reconstruct those files before 
continuing.
   - **Fixed in `e8f181cdf51e6bc9d70781d301cdd6ac16976202`.** Replay the 
preceding snapshot pieces without advancing the durable checkpoint, then 
continue from the saved request. A reconstruction failure retains the original 
uncertain request, and only that request may accept an idempotent conflict. 
Ordinary requests retain request-level resume; an acknowledged seal is not 
replayed. The checkpoint format is unchanged. Changes are limited to the CLI 
implementation and its test class.
   - **Local verification passed.** The regressions first reproduced the 
reviewed behavior: one offset-reset error, one failure showing a seal sent 
instead of a piece, and one passing acknowledged-seal boundary case. After the 
fix, all 30 reactor modules succeeded; Surefire XML confirms **77 tests, zero 
failures/errors/skips**: CLI 22, writer 30, archive reader 17, DataNode sink 8. 
This includes schema/config multi-file reconstruction, another interruption 
after an uncertain seal, and exact replay-permission assertions. Dependencies 
were built from clean source, then reused with matching upstream sources, 
profiles and toolchain; changed CLI sources were recompiled. Environment: 
Windows, JDK 17, Maven 3.8.8, English locale, one Maven thread, build cache 
disabled. Existing third-party dependencies were reused. Spotless apply with 
the integration-test profile and `git diff --check` passed. All newly added 
tests ran locally.
   - **Verification limits.** POSIX-only permission assertions, dual-cluster 
ITs, and full process-restart/leader-failover scenarios were not run in this 
round. Earlier results remain historical. The snapshot-resume tests use real 
wire payloads and a simulated receiver with empty staging, offset checks and 
final-byte assertions.
   - **Remote CI is separate.** At 03:24:26 UTC, the reviewed SHA had 41 
checks: 29 successful, 1 failed ([SonarCloud Code 
Analysis](https://github.com/apache/iotdb/runs/126359073560), new-code 
reliability gate), and 11 unfinished. At 03:28:26 UTC, the final SHA had 27 
visible checks, all unfinished (26 running, 1 queued); combined commit status 
was pending. No final-head CI pass is claimed.
   - **Final state confirmed:** local HEAD = `apache/iotdb` source branch 
`feat/pipe-logical-backup-sink` = PR #18623 head = 
**`e8f181cdf51e6bc9d70781d301cdd6ac16976202`**. The explicit source-branch push 
succeeded and the PR remains **Open/Draft**. This round found a behavioral 
issue, so a fresh independent session is required before marking it ready. No 
second review round was performed here.
   


-- 
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