[
https://issues.apache.org/jira/browse/HADOOP-13028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15279905#comment-15279905
]
Steve Loughran commented on HADOOP-13028:
-----------------------------------------
Because one place I'm using this to look at the logs and see how to tune the
performance is in spark code which doesn't have access to those internals and
is built against Hadoop 2.6.x anyway. It lets me have code which can be run
with -Dhadoop.version=2.7.1 and -Dhadoop.version=2.8.0-SNAPSHOT, I can not only
measure the duration in the spark code itself, I can see the logged info and
see what's been happening —where things can be improved futher.
We cannot do this if the way to log this data is via a class which is package
private and in Hadoop 2.8+ only. As requested, I've scoped that statistics
class so that the only way to get at it is to inject code into the
org.apache.hadoop.fs.s3a package. Do you really, really, want me to do that in
spark code? And use introspection to get at a class it can't compile against.
Please, give me the string: it'll be better for all of us. As and when your
colleagues sit down to look at Parquet performance on S3, they'll appreciate it.
> add low level counter metrics for S3A; use in read performance tests
> --------------------------------------------------------------------
>
> Key: HADOOP-13028
> URL: https://issues.apache.org/jira/browse/HADOOP-13028
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3, metrics
> Affects Versions: 2.8.0
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Attachments: HADOOP-13028-001.patch, HADOOP-13028-002.patch,
> HADOOP-13028-004.patch, HADOOP-13028-005.patch, HADOOP-13028-006.patch,
> HADOOP-13028-007.patch, HADOOP-13028-008.patch, HADOOP-13028-009.patch,
> HADOOP-13028-branch-2-008.patch, HADOOP-13028-branch-2-009.patch,
> HADOOP-13028-branch-2-010.patch, HADOOP-13028-branch-2-011.patch,
> org.apache.hadoop.fs.s3a.scale.TestS3AInputStreamPerformance-output.txt,
> org.apache.hadoop.fs.s3a.scale.TestS3AInputStreamPerformance-output.txt
>
>
> against S3 (and other object stores), opening connections can be expensive,
> closing connections may be expensive (a sign of a regression).
> S3A FS and individual input streams should have counters of the # of
> open/close/failure+reconnect operations, timers of how long things take. This
> can be used downstream to measure efficiency of the code (how often
> connections are being made), connection reliability, etc.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
