GitHub user bOOm-X opened a pull request: https://github.com/apache/spark/pull/18004
[SPARK-18838][CORE] Introduce blocking strategy for LiveListener ## What changes were proposed in this pull request? When the queue of the LiveListener is full, events are dropped in an arbitrary fashion. But some of them can be crucial for the good execution of the job. The job doesn t in general fail immediately, but presents incoherent state before dying. In this pull request: the queue implementation was refactored to increase a bit the performances (reduce object allocation & decrease lock usage) A "waiting for space" strategy is introduced and can be used instead of the default and current "dropping" strategy. The new strategy even if it does not resolve the "real" problem (the small dequeing rate) brings a lot more stability to the jobs with a huge event production rate (i.e jobs with many partitions). ## How was this patch tested? Existing unit tests have been enriched to run on both strategy. + manual tests have been run on the cluster You can merge this pull request into a Git repository by running: $ git pull https://github.com/bOOm-X/spark master Alternatively you can review and apply these changes as the patch at: https://github.com/apache/spark/pull/18004.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #18004 ---- commit 445f502bf937d57ffbbb05afd4251f957dc008a6 Author: bOOm-X <antoine.pr...@gmail.com> Date: 2017-05-10T18:16:51Z [SPARK-18838][CORE] Introduce blocking strategy for LiveListener ## What changes were proposed in this pull request? When the queue of the LiveListener is full, events are dropped in an arbitrary fashion. But some of them can be crucial for the good execution of the job. the job doesn t in general fail immediately, but presents incoherent state before dying. In this changelist: the queue implementation to increase a bit the performances (reduce object allocation & decrease lock usage) A "waiting for space" strategy is introduced and can be used instead of the default and current "dropping" strategy. The new strategy even if it does not resolve the "real" problem (the small dequeing rate) bring a lot more stability to the jobs with a huge event production rate (i.e jobs with many partitions). ## How was this patch tested? Existing unit tests have been enriched to run on both strategy. + manual tests have been run on the cluster ---- --- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org