Rangsh opened a new issue, #12266: URL: https://github.com/apache/seatunnel/issues/12266
## Motivation Follow-up from #12173 / #12063 review (SEZ9 F3). `IMapJobGrowthBenchmarkWorkload` currently samples MapStore durability by reloading **one representative key** (plus asserting the rest of the growth batch remains resident). That is intentional: `FileMapStore.loadAll` always replays the full append-only WAL into heap, and a mid-trial full-batch `loadAll` under `initialStoredJobCount=1000` previously OOM'd diagnostic / JMH forks. ## Proposed follow-up - Widen the durability sample to a full-batch `loadAll` (or an equivalent per-key durable round-trip) when it can be done without blowing the fork heap budget. - Keep the sample **outside** measured SingleShot time (iteration / trial tear-down only). - Preserve the existing first empty-pressure iteration + trial tear-down sampling windows unless a safer mid-trial strategy is found. ## Acceptance criteria - [ ] Full growth batch is durability-verified from MapStore (not only one representative key + resident checks) - [ ] No OOM under `initialStoredJobCount=1000` for normal Benchmarks / Diagnostics forks - [ ] Measured growth scores remain free of full-WAL reload cost between SingleShot samples ## Related - PR: https://github.com/apache/seatunnel/pull/12173 - Parent: https://github.com/apache/seatunnel/issues/12063 -- 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]
