From 3ddf051ee7256f642f8a17768d161c7b5f55c7e1 Mon Sep 17 00:00:00 2001 From: ehnalis Date: Wed, 20 May 2015 08:27:39 -0500 Subject: [SPARK-7533] [YARN] Decrease spacing between AM-RM heartbeats. Added faster RM-heartbeats on pending container allocations with multiplicative back-off. Also updated related documentations. Author: ehnalis Closes #6082 from ehnalis/yarn and squashes the following commits: a1d2101 [ehnalis] MIss-spell fixed. 90f8ba4 [ehnalis] Changed default HB values. 6120295 [ehnalis] Removed the bug, when allocation heartbeat would not start from initial value. 08bac63 [ehnalis] Refined style, grammar, removed duplicated code. 073d283 [ehnalis] [SPARK-7533] [YARN] Decrease spacing between AM-RM heartbeats. d4408c9 [ehnalis] [SPARK-7533] [YARN] Decrease spacing between AM-RM heartbeats. --- docs/running-on-yarn.md | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'docs/running-on-yarn.md') diff --git a/docs/running-on-yarn.md b/docs/running-on-yarn.md index 51c1339165..9d55f435e8 100644 --- a/docs/running-on-yarn.md +++ b/docs/running-on-yarn.md @@ -71,9 +71,22 @@ Most of the configs are the same for Spark on YARN as for other deployment modes spark.yarn.scheduler.heartbeat.interval-ms - 5000 + 3000 The interval in ms in which the Spark application master heartbeats into the YARN ResourceManager. + The value is capped at half the value of YARN's configuration for the expiry interval + (yarn.am.liveness-monitor.expiry-interval-ms). + + + + spark.yarn.scheduler.initial-allocation.interval + 200ms + + The initial interval in which the Spark application master eagerly heartbeats to the YARN ResourceManager + when there are pending container allocation requests. It should be no larger than + spark.yarn.scheduler.heartbeat.interval-ms. The allocation interval will doubled on + successive eager heartbeats if pending containers still exist, until + spark.yarn.scheduler.heartbeat.interval-ms is reached. -- cgit v1.2.3