1996fanrui commented on PR #27659: URL: https://github.com/apache/flink/pull/27659#issuecomment-5713716260
Hey @Shekharrajak @RocMarshal , I am upgrading flink from 2.0 to 2.3, and found the CI get slower. Seems caused by this pr. Heads-up: this 30s `REQUIREMENTS_CHECK_DELAY` override makes `EventTimeWindowCheckpointingITCase` ~6.5× slower — every parameterized case idles ~30s before its job is scheduled (default is 50ms). The de-flake is effective, but the ~30s is paid on every invocation on the happy path. Per-case seconds (same code/env, only difference = this line; `testSlidingTimeWindow` disabled): ### With the 30s line | Method | MEM | FILE | ROCKSDB_FULL | ROCKSDB_INCR | ROCKSDB_INCR_ZK | FORST_INCR | |---|---|---|---|---|---|---| | testTumblingTimeWindow | 34.6 | 34.5 | 37.1 | 35.1 | 34.1 | 34.5 | | testTumblingTimeWindowWithKVStateMinMaxParallelism | 38.4 | 34.7 | 35.5 | 33.8 | 35.3 | 35.0 | | testTumblingTimeWindowWithKVStateMaxMaxParallelism | 36.8 | 37.6 | 35.3 | 33.8 | 35.8 | 33.7 | | testPreAggregatedTumblingTimeWindow | 33.7 | 33.4 | 34.7 | 34.9 | 34.1 | 34.6 | | testPreAggregatedSlidingTimeWindow | 34.8 | 34.7 | 37.6 | 36.9 | 38.7 | 36.6 | | **column total** | 178.2 | 174.9 | 180.1 | 174.5 | 178.0 | 174.4 | ### Without the line (default 50ms) | Method | MEM | FILE | ROCKSDB_FULL | ROCKSDB_INCR | ROCKSDB_INCR_ZK | FORST_INCR | |---|---|---|---|---|---|---| | testTumblingTimeWindow | 4.4 | 4.5 | 4.8 | 3.8 | 5.0 | 4.7 | | testTumblingTimeWindowWithKVStateMinMaxParallelism | 9.0 | 4.8 | 6.3 | 3.8 | 5.4 | 5.1 | | testTumblingTimeWindowWithKVStateMaxMaxParallelism | 8.0 | 7.5 | 4.9 | 5.1 | 4.2 | 4.7 | | testPreAggregatedTumblingTimeWindow | 4.8 | 3.6 | 3.6 | 4.8 | 5.5 | 3.8 | | testPreAggregatedSlidingTimeWindow | 4.9 | 5.8 | 6.7 | 7.8 | 8.6 | 7.7 | | **column total** | 31.1 | 26.1 | 26.2 | 25.2 | 28.7 | 26.0 | ~30s slower per case → class portion ~1060s → ~163s. All cases pass either way. -- 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]
