[
https://issues.apache.org/jira/browse/HADOOP-13230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18011692#comment-18011692
]
ASF GitHub Bot commented on HADOOP-13230:
-----------------------------------------
liapengpony commented on code in PR #2149:
URL: https://github.com/apache/hadoop/pull/2149#discussion_r2250007813
##########
hadoop-tools/hadoop-aws/src/main/java/org/apache/hadoop/fs/s3a/S3AFileSystem.java:
##########
@@ -4086,25 +4175,41 @@ public boolean exists(Path f) throws IOException {
}
/**
- * Override superclass so as to add statistic collection.
+ * Optimized probe for a path referencing a dir.
+ * Even though it is optimized to a single HEAD, applications
+ * should not over-use this method...it is all too common.
* {@inheritDoc}
*/
@Override
@SuppressWarnings("deprecation")
public boolean isDirectory(Path f) throws IOException {
Review Comment:
@steveloughran it looks the change to this function was meant to optimize
performance, but I am experiencing performance regression when upgrading spark
version from 3.1.2 to 3.5.1, and I found it caused by this change.
When I do a spark.read.parquet('s3a://path/to/1.parquet', ...,
's3a://path/to/10000.parquet') before this change, ONLY HEAD requests are sent
to build the DataFrame. However, after this change, LIST requests are sent,
which is significantly slower as I am reading from quite a lot of parquets.
The docstring "it is optimized to a single HEAD" also confuses me because
StatusProbeEnum.DIRECTORIES is just an alias for StatusProbeEnum.LIST_ONLY.
Am I missing anything here?
> S3A to optionally retain directory markers
> ------------------------------------------
>
> Key: HADOOP-13230
> URL: https://issues.apache.org/jira/browse/HADOOP-13230
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 2.9.0
> Reporter: Aaron Fabbri
> Assignee: Steve Loughran
> Priority: Major
> Labels: pull-request-available
> Fix For: 3.3.1
>
> Attachments: 2020-02-Fixing the S3A directory marker problem.pdf
>
> Time Spent: 50m
> Remaining Estimate: 0h
>
> Users of s3a may not realize that, in some cases, it does not interoperate
> well with other s3 tools, such as the AWS CLI. (See HIVE-13778, IMPALA-3558).
> Specifically, if a user:
> - Creates an empty directory with hadoop fs -mkdir s3a://bucket/path
> - Copies data into that directory via another tool, i.e. aws cli.
> - Tries to access the data in that directory with any Hadoop software.
> Then the last step fails because the fake empty directory blob that s3a wrote
> in the first step, causes s3a (listStatus() etc.) to continue to treat that
> directory as empty, even though the second step was supposed to populate the
> directory with data.
> I wanted to document this fact for users. We may mark this as not-fix, "by
> design".. May also be interesting to brainstorm solutions and/or a config
> option to change the behavior if folks care.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
