Sam Whittle created BEAM-12740:
----------------------------------
Summary: Reduce and backoff GCS metadata operations when writing
to GCS files
Key: BEAM-12740
URL: https://issues.apache.org/jira/browse/BEAM-12740
Project: Beam
Issue Type: Bug
Components: io-java-gcp
Reporter: Sam Whittle
Assignee: Sam Whittle
When issuing GCS operations affecting metadata (ie file-level operations not
read/write operations), GCS may return errors indicating backoff. See
https://cloud.google.com/storage/docs/request-rate#ramp-up
If such errors are encountered, currently the exception is not handled by
GcsUtil.java and is propagated, causing retries and backoff of all operations
at a higher level. Instead we should backoff and retry only such files that
require it.
Additionally FileBasedSink issues deletes for files that have been renamed.
The rename itself should take care of removing the original file and thus we
can reduce some metadata operations.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
