mqliang edited a comment on pull request #6710: URL: https://github.com/apache/incubator-pinot/pull/6710#issuecomment-809727115
@siddharthteotia @mcvsubbu Forced updated to delete all stale changes. PR is ready for review. > We can emit both costs separately. So have 2 server gauges. Similarly, log them in the QueryScheduler separately. But the serialized cost in the DataTable should be a single value (sum total of both exec cpu time cost and serialization cpu time cost) This implementation is cumbersome -- either need serialize data table twice, or need hack the bytes to do some content replacement: We first add `executionThreadCpuTimeNs` into metadata when execute query, then add `serializationCpuTimeNs` into metadata when serialize data table, log and emit emit both costs separately. Then sum them together, remove them from metadata and add a `totalThreadCpuTimeNs` into metadata, then we need serialize datatable again to send back to broker. My implementation log and emit a sum metric only, which is much simpler and easy to read: in `datatable.toBytes()`, before serialize metadata, update the value of "executionThreadCpuTimeNs" to account data table serialization time. -- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org For additional commands, e-mail: commits-h...@pinot.apache.org
