steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-866696567
+ maybe I could test this by setting the sysprop `aws.region` to something
invalid. If the region resolution is going through the chain then this would
get picked up ahead
steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-866694906
One thought here: would you ever want the s3a connector to fall back to that
bundled region lookup sequence?
I'm wondering in particular if it makes a difference in
steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-866183472
tested with `-Dparallel-tests -DtestsThreadCount=7 -Dmarkers=keep -Dscale
-Ds3guard -Ddynamo`
transient read buffer underflow failure; one extra S3Guard write than
steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-865965433
Test run with `-Dparallel-tests -DtestsThreadCount=7 -Dmarkers=keep` -no
failures!
rerunning with dynamo & scale
--
This is an automated message from the Apache
steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-865943337
+ No new tests here. The only way to test would be to destroy ~/.aws/config
and & skip if the env vars/syssprops resolved or running in EC2. If we could
set the resolve
steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-865941502
## `cloudstore regions`
This verifies that the local system is set up to fail without the patch.
```
> bin/hadoop jar $CLOUDSTORE regions
steveloughran commented on pull request #3133:
URL: https://github.com/apache/hadoop/pull/3133#issuecomment-865940542
Testing in progress, in a setup where
* the failure mode is replicated
* fs.s3a.endpoint is not set globally or for the test bucket.
--
This is an automated