Okada Haruki

@stanislavkozlovski Hi, could you take a look?

Hi, thanks for reporting the issue. Though your use-case sounds not so common, definitely it's hard to achieve in current Decaton features. > Separate decaton.max.pending.records to 2 configurations This approach...

@thachlp Sorry for the late reply. Thank you for your interest! However this issue may likely be solved by https://github.com/line/decaton/pull/224

Thanks for reporting the issue. I have a question about the situation. > Finally, we solved this by keeping the deferred completion for the smallest offset for each topic partition...

related: https://github.com/line/centraldogma/issues/151

Seems it still happens https://github.com/line/decaton/actions/runs/4508725224/jobs/7937698592 ``` com.linecorp.decaton.centraldogma.CentralDogmaPropertySupplierIntegrationTest > testCDIntegration FAILED java.util.concurrent.CompletionException: com.linecorp.armeria.client.UnprocessedRequestException: com.linecorp.armeria.client.endpoint.EndpointSelectionTimeoutException: Failed to select within 3200 ms an endpoint from: HealthCheckedEndpointGroup{endpoints=[Endpoint{127.0.0.1:42919, weight=1000}], numEndpoints=1, candidates=[Endpoint{127.0.0.1:42919, weight=1000}], numCandidates=1, selectionStrategy=class com.linecorp.armeria.client.endpoint.WeightedRoundRobinStrategy,...

Thanks for reporting the issue. However, I prefer to keep WARN level for these commit failures because: - Though Decaton doesn't assume commit-failure is fatal as commit is attempted soon...

I think you can try [EvaluationFilter](https://logback.qos.ch/manual/filters.html#evaluatorFilter) and set the expression like ``` return throwable instanceof org.apache.kafka.clients.consumer.RetriableCommitFailedException; ```

> I am currently reworking some infrastructure to allow duckdb-wasm to build loadable extensions. > Once that lands I will see how this PR can fit That sounds great, thanks!

- Bad performance on Safari - Broken with large recording on Chrome