Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/9182#discussion_r43376459
--- Diff:
yarn/src/main/scala/org/apache/spark/scheduler/cluster/YarnSchedulerBackend.scala
---
@@ -51,6 +51,41 @@ private[spark] abstract class YarnSchedulerBackend(
private implicit val askTimeout = RpcUtils.askRpcTimeout(sc.conf)
+ /** Application ID. Must be set by a subclass before starting the
service */
+ private var appId: ApplicationId = null
+
+ /** Attempt ID. This is unset for client-mode schedulers */
+ private var attemptId: Option[ApplicationAttemptId] = None
+
+ /** Scheduler extension services */
+ private val services: SchedulerExtensionServices = new
SchedulerExtensionServices()
+
+ /**
+ * Bind to YARN. This *must* be done before calling [[start()]].
+ *
+ * @param appId YARN application ID
+ * @param attemptId Optional YARN attempt ID
+ */
+ protected def bindToYarn(appId: ApplicationId, attemptId:
Option[ApplicationAttemptId]): Unit = {
+ this.appId = appId
+ this.attemptId = attemptId
+ }
+
+ override def start() {
+ require(appId != null, "application ID unset")
+ val binding = SchedulerExtensionServiceBinding(sc, appId, attemptId)
+ services.start(binding)
--- End diff --
string parsing has proven fairly brittle in the past; the move from single
to multiple attempts broke all apps trying to do it across versions (i.e. a
hadoop 2.2 parser in a 2.5 cluster). Unless you want to base-64 encode the
protobuf representation, I'd avoid that.
---
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
