HeartSaVioR commented on issue #26690: [SPARK-30057][DOCS]Add a statement of platforms Spark runs on URL: https://github.com/apache/spark/pull/26690#issuecomment-558994285 Just 2 cents as I'm not a committer nor a PMC member of Apache Spark; I think there's still a non-trivial gap between "all UTs pass" and "guaranteed to work". Even Spark runs around 3~4hrs of UTs for every PRs as well as master branch, the benefit of UTs is more alike avoiding regressions and it doesn't prevent any further bugs including correctness and dataloss. IMHO, the core circumstances of "guaranteed to work" are following: 1) Have end-to-end system tests running in such environment as well. (If it includes chaos monkey then much better.) I believe some of vendors are running their own system tests in their own environments. 2) Core staffs of project are familiar with such environment and able to deal with any env. specific issues. 3) End users have been running various workloads so far. I agree that's egg vs chicken problem for ARM support, but at least we may want to see couple of references for (pseudo) real workloads. but everyone may have different opinions on the circumstances. Someone may have even higher bar on this, someone may have lower bar - so please just treat this as 2 cents, as I mentioned above.
---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
