jt2594838 opened a new pull request, #19242:
URL: https://github.com/apache/iotdb/pull/19242

   Compatible integration-test methods repeatedly start identical clusters, 
create clients for individual writes/flushes, and wait through fixed polling or 
observation delays. This change shares class-owned clusters, isolates or resets 
scenario data and metadata, reuses Sessions, and batches fixture writes. The 
DataSink fixture reduces repeated Tablet writes from 125 to 12 per generator 
while retaining four write/flush rounds, both databases and all five table 
shapes. Query fixtures retain their input values; recovery cases retain their 
required restarts.
   
   Use immediate, bounded condition polling for readiness, quota/CQ/TTL 
visibility and asynchronous cleanup. Data-only Pipe cases wait for a full-flush 
completion barrier and then check the exact receiver results; negative 
permission/filter observations retain their required observation windows. 
Restore dynamic configuration between shared-cluster methods, bound the LOAD 
fixture's Region budget, and make acknowledged-write durability explicit with 
SYNC WAL in restart fixtures. Cleanup tolerates only an already-deleted 
database when an in-flight metadata DROP wins the race. Add opt-in ConfigNode 
startup tracing and client-side phase timings to explain remaining costs. All 
changes are in `integration-test`.
   
   Measured once per version, serially on the same Windows machine at 
`592d6635027c0a4dea0fa461f2bb6f0c56bc15e8`, before porting to current master. 
Class durations include fixture setup/cleanup. Of 165 class/profile 
combinations, 156 passed on both versions with identical method/skip sets and 
valid execution evidence; nine with baseline failures are excluded from the 
savings:
   
   | Comparable tests | Before (s) | After (s) | Reduction |
   | --- | ---: | ---: | ---: |
   | Pipe IT | 8,826.987 | 2,551.928 | 71.1% |
   | Other IT | 7,576.007 | 2,505.097 | 66.9% |
   | Total | 16,402.994 | 5,057.025 | 69.2% |
   
   The optimized original-revision run recorded 1,417 passed and 73 skipped 
with zero failures/errors. These are single-pair local measurements, not CI 
timing estimates. The final patch preserves upstream master regressions and 
adds the target branch's credential/namespace adaptations; current-master 
validation is reported separately below.
   
   Validation:
   
   - Spotless, IT discovery/shard coverage, patch application to frozen master, 
new-file license headers, and en/zh message-key/placeholder parity checked.
   - Default and Chinese full-reactor test compilation passed (51 modules per 
locale).
   - On `7b407ec00292`: 165 class/profile combinations, 1454 passed, 71 
skipped, 0 failures and 0 errors. The complete first run exposed fixture 
isolation/durability issues; after fixing them, affected classes and 
descendants of the repaired shared fixture were rerun in full. Method/category 
coverage was checked against the source, with no new Ignore annotations.
   


-- 
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