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.
-
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_STORE
format. 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