Replies: 4 comments 3 replies
-
|
I also tested by setting It seems like the empty directory markers are what's causing the issue: |
Beta Was this translation helpful? Give feedback.
-
|
@mcgrawia A couple of S3A compatibility issues were fixed since 1.4.0, please see HDDS-8423. Especially HDDS-10570 and HDDS-10572 seem to be related to your woes. Can you please try building Ozone from source ( |
Beta Was this translation helpful? Give feedback.
-
The A tentative workaround is to disable shallow listing by setting the property |
Beta Was this translation helpful? Give feedback.
-
|
Hello @mcgrawia, maybe this instruction help you https://docs.cloudera.com/cdp-private-cloud-base/7.1.9/ozone-storing-data/topics/ozone-config-spark-s3a.html |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi,
I'm trying to use Ozone for the first time and running into issues with Spark / S3 compatibility and was hoping someone here may be able to help me. Thanks in advance!
Summary of problem
My Spark job attempts to overwrite a Spark ModelPipeline object:
This jobs fails with the following errors:
When I list the contents of the bucket, I see a lot of empty files, which seem to be preventing the MultiObjectDelete command from succeeding:
Troubleshooting steps
I've also tried creating my bucket using the
--layout OBJECT_STOREformat. This seems to cause a different issue. I no longer see the error above, but my spark job eventually fails with this log:When I look into the S3G pod, I see the following logs:
Looking using the AWS CLI, I can see there seems to be a directory marker file in the bucket:
But when I attempt to get this file, the AWS CLI fails:
Some Background / Setup Details
I'm running Ozone via the Ozone helm chart in Kubernetes with default values, using version 1.4.0:
I created my bucket with the following commands:
and set the S3G domain in my
ozone-site.xml:My current Spark S3A settings are the following:
Additionally, I have also tried setting
spark.hadoop.fs.s3a.directory.marker.retention=delete, and it did not seem to change anything.Thanks for the help 🙏
Beta Was this translation helpful? Give feedback.
All reactions