[
https://issues.apache.org/jira/browse/HADOOP-17415?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18039333#comment-18039333
]
ASF GitHub Bot commented on HADOOP-17415:
-----------------------------------------
monthonk opened a new pull request, #3939:
URL: https://github.com/apache/hadoop/pull/3939
### Description of PR
As part of all the openFile work, knowing full length of an object allows
for a HEAD to be skipped. But: code knowing only the splits don't know the
final length of the file.
If the content-range header is used, then as soon as a single GET is
initiated against an object, if the field is returned then we can update the
length of the S3A stream to its real/final length
* Skip file status probe on openFile. Content length will be updated on
first read.
* As a side effect, modification time, etag and version id also be unknown
as there is no probe on openFile.
* Allow content length to be negative value, it means content length is
unknown.
* On read file, use information from Content-Range header to update content
length.
* If out of range exception occurs on read, ActualObjectSize from exception
details can be used to update content length.
### How was this patch tested?
Tested in `eu-west-1` with `mvn -Dparallel-tests -DtestsThreadCount=16 clean
verify` (AP tests failed from previous SDK upgrade)
```
[INFO] Results:
[INFO]
[ERROR] Errors:
[ERROR]
ITestS3ABucketExistence.testAccessPointProbingV2:171->expectUnknownStore:103->lambda$testAccessPointProbingV2$12:172
» IllegalArgument
[ERROR]
ITestS3ABucketExistence.testAccessPointRequired:188->expectUnknownStore:103->lambda$testAccessPointRequired$14:189
» IllegalArgument
[INFO]
[ERROR] Tests run: 1063, Failures: 0, Errors: 2, Skipped: 186
[INFO] Results:
[INFO]
[WARNING] Tests run: 108, Failures: 0, Errors: 0, Skipped: 68
```
### For code changes:
- [x] Does the title or this PR starts with the corresponding JIRA issue id
(e.g. 'HADOOP-17799. Your PR title ...')?
- [x] Object storage: have the integration tests been executed and the
endpoint declared according to the connector-specific documentation?
- [ ] If adding new dependencies to the code, are these dependencies
licensed in a way that is compatible for inclusion under [ASF
2.0](http://www.apache.org/legal/resolved.html#category-a)?
- [ ] If applicable, have you updated the `LICENSE`, `LICENSE-binary`,
`NOTICE-binary` files?
> Use S3 content-range header to update length of an object during reads
> ----------------------------------------------------------------------
>
> Key: HADOOP-17415
> URL: https://issues.apache.org/jira/browse/HADOOP-17415
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: fs/s3
> Affects Versions: 3.3.0
> Reporter: Steve Loughran
> Assignee: Monthon Klongklaew
> Priority: Major
> Labels: pull-request-available
> Time Spent: 50m
> Remaining Estimate: 0h
>
> As part of all the openFile work, knowing full length of an object allows for
> a HEAD to be skipped. But: code knowing only the splits don't know the final
> length of the file.
> If the content-range header is used, then as soon as a single GET is
> initiated against an object, if the field is returned then we can update the
> length of the S3A stream to its real/final length
> Also: when any input stream fails with an EOF exception, we can distinguish
> stream-interrupted from "no, too far"
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
