Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/6935#discussion_r47017107
--- Diff:
core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala ---
@@ -430,8 +519,54 @@ private[history] class FsHistoryProvider(conf:
SparkConf, clock: Clock)
}
newIterator.foreach(addIfAbsent)
oldIterator.foreach(addIfAbsent)
+ mergedApps
+ }
+
+
+ /**
+ * Scan through all incomplete applications, and check for any that have
changed.
+ *
+ * After the scan, if there were any updated attempts, [[applications]]
is updated
+ * with the new values.
+ *
+ * 1. No attempt to replay the application is made; this scan is a low
cost operation.
+ * 2. As this overwrites [[applications]] with a new value, it must not
run concurrently
+ */
+ private[history] def scanAndUpdateIncompleteAttemptInfo(): Unit = {
+ val now = System.currentTimeMillis();
+ val newAttempts: Iterable[FsApplicationAttemptInfo] = applications
+ .filter( e => !e._2.completed)
+ .flatMap { e =>
+ // build list of (false, attempt) or (true, attempt') values
+ e._2.attempts.flatMap { attempt: FsApplicationAttemptInfo =>
+ val path = new Path(logDir, attempt.logPath)
+ try {
+ val status = fs.getFileStatus(path)
+ val size = getLogSize(status).getOrElse(-1L)
+ val aS = attempt.fileSize
+ if (size > aS) {
--- End diff --
1. renamed
2. Not using timestamp, just size (and rm'd that unused
System.getTimeMillis()) flag. There's some interesting race conditions which
showed up in testing, which seemed fairly brittle to modtime changes. More
succinctly: I don't think HDFS and Posix work the same w.r.t modtime.
---
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]
