[GitHub] spark pull request #13830: [SPARK-16121] ListingFileCatalog does not list in...
Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/13830#discussion_r74684998
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/ListingFileCatalog.scala
---
@@ -73,21 +73,67 @@ class ListingFileCatalog(
cachedPartitionSpec = null
}
- protected def listLeafFiles(paths: Seq[Path]):
mutable.LinkedHashSet[FileStatus] = {
+ /**
+ * List leaf files of given paths. This method will submit a Spark job
to do parallel
+ * listing whenever there is a path having more files than the parallel
partition discovery
+ * discovery threshold.
+ */
+ protected[spark] def listLeafFiles(paths: Seq[Path]):
mutable.LinkedHashSet[FileStatus] = {
if (paths.length >=
sparkSession.sessionState.conf.parallelPartitionDiscoveryThreshold) {
HadoopFsRelation.listLeafFilesInParallel(paths, hadoopConf,
sparkSession)
} else {
+ // Right now, the number of paths is less than the value of
+ // parallelPartitionDiscoveryThreshold. So, we will list file
statues at the driver.
+ // If there is any child that has more files than the threshold, we
will use parallel
+ // listing.
+
// Dummy jobconf to get to the pathFilter defined in configuration
val jobConf = new JobConf(hadoopConf, this.getClass)
val pathFilter = FileInputFormat.getInputPathFilter(jobConf)
+
val statuses: Seq[FileStatus] = paths.flatMap { path =>
val fs = path.getFileSystem(hadoopConf)
logTrace(s"Listing $path on driver")
-Try {
- HadoopFsRelation.listLeafFiles(fs, fs.getFileStatus(path),
pathFilter)
-}.getOrElse(Array.empty[FileStatus])
+
+val childStatuses = {
+ // TODO: We need to avoid of using Try at here.
+ val stats =
Try(fs.listStatus(path)).getOrElse(Array.empty[FileStatus])
+ if (pathFilter != null) stats.filter(f =>
pathFilter.accept(f.getPath)) else stats
+}
+
+childStatuses.map {
+ case f: LocatedFileStatus => f
+
+ // NOTE:
+ //
+ // - Although S3/S3A/S3N file system can be quite slow for
remote file metadata
+ // operations, calling `getFileBlockLocations` does no harm
here since these file system
+ // implementations don't actually issue RPC for this method.
+ //
--- End diff --
Maybe just say "Location aware object stores may trigger queries of the
object store here" without being specific about the store. As an example:
`org.apache.hadoop.fs.swift.snative.SwiftNativeFileSystem` does this.
---
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
[GitHub] spark pull request #13830: [SPARK-16121] ListingFileCatalog does not list in...
Github user yhuai commented on a diff in the pull request:
https://github.com/apache/spark/pull/13830#discussion_r72515446
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/ListingFileCatalog.scala
---
@@ -73,21 +73,67 @@ class ListingFileCatalog(
cachedPartitionSpec = null
}
- protected def listLeafFiles(paths: Seq[Path]):
mutable.LinkedHashSet[FileStatus] = {
+ /**
+ * List leaf files of given paths. This method will submit a Spark job
to do parallel
+ * listing whenever there is a path having more files than the parallel
partition discovery
+ * discovery threshold.
+ */
+ protected[spark] def listLeafFiles(paths: Seq[Path]):
mutable.LinkedHashSet[FileStatus] = {
if (paths.length >=
sparkSession.sessionState.conf.parallelPartitionDiscoveryThreshold) {
HadoopFsRelation.listLeafFilesInParallel(paths, hadoopConf,
sparkSession)
} else {
+ // Right now, the number of paths is less than the value of
+ // parallelPartitionDiscoveryThreshold. So, we will list file
statues at the driver.
+ // If there is any child that has more files than the threshold, we
will use parallel
+ // listing.
+
// Dummy jobconf to get to the pathFilter defined in configuration
val jobConf = new JobConf(hadoopConf, this.getClass)
val pathFilter = FileInputFormat.getInputPathFilter(jobConf)
+
val statuses: Seq[FileStatus] = paths.flatMap { path =>
val fs = path.getFileSystem(hadoopConf)
logTrace(s"Listing $path on driver")
-Try {
- HadoopFsRelation.listLeafFiles(fs, fs.getFileStatus(path),
pathFilter)
-}.getOrElse(Array.empty[FileStatus])
+
+val childStatuses = {
+ // TODO: We need to avoid of using Try at here.
+ val stats =
Try(fs.listStatus(path)).getOrElse(Array.empty[FileStatus])
+ if (pathFilter != null) stats.filter(f =>
pathFilter.accept(f.getPath)) else stats
+}
+
+childStatuses.map {
+ case f: LocatedFileStatus => f
+
+ // NOTE:
+ //
+ // - Although S3/S3A/S3N file system can be quite slow for
remote file metadata
+ // operations, calling `getFileBlockLocations` does no harm
here since these file system
+ // implementations don't actually issue RPC for this method.
+ //
--- End diff --
Thank you for pointing out that. Can you also provide some pointers? We can
put those pointers in the comment.
---
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
[GitHub] spark pull request #13830: [SPARK-16121] ListingFileCatalog does not list in...
Github user steveloughran commented on a diff in the pull request:
https://github.com/apache/spark/pull/13830#discussion_r72242750
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/ListingFileCatalog.scala
---
@@ -73,21 +73,67 @@ class ListingFileCatalog(
cachedPartitionSpec = null
}
- protected def listLeafFiles(paths: Seq[Path]):
mutable.LinkedHashSet[FileStatus] = {
+ /**
+ * List leaf files of given paths. This method will submit a Spark job
to do parallel
+ * listing whenever there is a path having more files than the parallel
partition discovery
+ * discovery threshold.
+ */
+ protected[spark] def listLeafFiles(paths: Seq[Path]):
mutable.LinkedHashSet[FileStatus] = {
if (paths.length >=
sparkSession.sessionState.conf.parallelPartitionDiscoveryThreshold) {
HadoopFsRelation.listLeafFilesInParallel(paths, hadoopConf,
sparkSession)
} else {
+ // Right now, the number of paths is less than the value of
+ // parallelPartitionDiscoveryThreshold. So, we will list file
statues at the driver.
+ // If there is any child that has more files than the threshold, we
will use parallel
+ // listing.
+
// Dummy jobconf to get to the pathFilter defined in configuration
val jobConf = new JobConf(hadoopConf, this.getClass)
val pathFilter = FileInputFormat.getInputPathFilter(jobConf)
+
val statuses: Seq[FileStatus] = paths.flatMap { path =>
val fs = path.getFileSystem(hadoopConf)
logTrace(s"Listing $path on driver")
-Try {
- HadoopFsRelation.listLeafFiles(fs, fs.getFileStatus(path),
pathFilter)
-}.getOrElse(Array.empty[FileStatus])
+
+val childStatuses = {
+ // TODO: We need to avoid of using Try at here.
+ val stats =
Try(fs.listStatus(path)).getOrElse(Array.empty[FileStatus])
+ if (pathFilter != null) stats.filter(f =>
pathFilter.accept(f.getPath)) else stats
+}
+
+childStatuses.map {
+ case f: LocatedFileStatus => f
+
+ // NOTE:
+ //
+ // - Although S3/S3A/S3N file system can be quite slow for
remote file metadata
+ // operations, calling `getFileBlockLocations` does no harm
here since these file system
+ // implementations don't actually issue RPC for this method.
+ //
--- End diff --
while you are correct about s3a, the openstack swift client does do RPC
here, as swift instances can exhibit locality, as does Azure. This comment
isn't actually correct.
---
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
[GitHub] spark pull request #13830: [SPARK-16121] ListingFileCatalog does not list in...
Github user asfgit closed the pull request at: https://github.com/apache/spark/pull/13830 --- 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
[GitHub] spark pull request #13830: [SPARK-16121] ListingFileCatalog does not list in...
GitHub user yhuai opened a pull request: https://github.com/apache/spark/pull/13830 [SPARK-16121] ListingFileCatalog does not list in parallel anymore ## What changes were proposed in this pull request? Seems the fix of SPARK-14959 breaks the parallel partitioning discovery. This PR fixes the problem ## How was this patch tested? Will manually test it. (This PR also adds a proper test for SPARK-14959) You can merge this pull request into a Git repository by running: $ git pull https://github.com/yhuai/spark SPARK-16121 Alternatively you can review and apply these changes as the patch at: https://github.com/apache/spark/pull/13830.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 #13830 commit 03b4497524bb62c8a87dd3b3775006ad162a1270 Author: Yin Huai Date: 2016-06-21T23:53:38Z [SPARK-16121] ListingFileCatalog does not list in parallel anymore This PR also provides an alternative fix for https://issues.apache.org/jira/browse/SPARK-14959 commit dff33388c15fcf5ebefeb52a4f67b6fd930f7d2c Author: Yin Huai Date: 2016-06-21T23:56:44Z Merge remote-tracking branch 'upstream/master' into SPARK-16121 --- 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
