Github user srowen commented on a diff in the pull request:
https://github.com/apache/spark/pull/5236#discussion_r27792094
--- Diff: core/src/main/scala/org/apache/spark/HeartbeatReceiver.scala ---
@@ -62,14 +62,15 @@ private[spark] class HeartbeatReceiver(sc: SparkContext)
// "spark.network.timeout" uses "seconds", while
`spark.storage.blockManagerSlaveTimeoutMs` uses
// "milliseconds"
- private val executorTimeoutMs =
sc.conf.getOption("spark.network.timeout").map(_.toLong * 1000).
- getOrElse(sc.conf.getLong("spark.storage.blockManagerSlaveTimeoutMs",
120000))
+ private val executorTimeoutMs = Utils.timeStringAsMs(
+ sc.conf.get("spark.storage.blockManagerSlaveTimeoutMs",
+ s"${Utils.timeStringAsS(sc.conf.get("spark.network.timeout",
"120s"))}s"))
// "spark.network.timeoutInterval" uses "seconds", while
// "spark.storage.blockManagerTimeoutIntervalMs" uses "milliseconds"
- private val checkTimeoutIntervalMs =
- sc.conf.getOption("spark.network.timeoutInterval").map(_.toLong *
1000).
-
getOrElse(sc.conf.getLong("spark.storage.blockManagerTimeoutIntervalMs", 60000))
+ private val checkTimeoutIntervalMs = Utils.timeStringAsMs(
+ sc.conf.get("spark.storage.blockManagerTimeoutIntervalMs",
--- End diff --
Am I reading this wrong or does this flip the order of priority for these
two properties? Looks like `spark.network.timeoutInterval` should take
precedence? but now it's the fallback default. If that's so, this might be
worth unpacking a bit with intermediate values anyway as it's a little hard to
read the logic
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
