Dylan Wong created SPARK-54290:
----------------------------------
Summary: Skip checksum creation if path already exists without
checksum
Key: SPARK-54290
URL: https://issues.apache.org/jira/browse/SPARK-54290
Project: Spark
Issue Type: Task
Components: Structured Streaming
Affects Versions: 4.2.0
Reporter: Dylan Wong
Consider the case using STATE_STORE_CHECKPOINT_FORMAT_VERSION = 1 when a batch
fails but state files are written. If on the next run, we try to upload both a
new state file and a file checksum, the file could fail to be uploaded but the
file checksum is uploaded successfully. This would lead to a situation where
the old file could be loaded and compared with the new file checksum, which
would fail the checksum verification. This issue does not happen when
STATE_STORE_CHECKPOINT_FORMAT_VERSION = 2 since each batch run unique ids will
be created.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
