Ufuk Celebi created FLINK-7676:
----------------------------------
Summary: ContinuousFileMonitoringFunction fails with
GoogleHadoopFileSystem
Key: FLINK-7676
URL: https://issues.apache.org/jira/browse/FLINK-7676
Project: Flink
Issue Type: Bug
Components: Streaming Connectors
Reporter: Ufuk Celebi
Priority: Minor
The following check in ContinuousFileMonitoringFunction fails when running
against a file in Google Cloud Storage:
{code}
Path p = new Path(path);
FileSystem fileSystem = FileSystem.get(p.toUri());
if (fileSystem.exists(p)) {
throw new FileNotFoundException("The provided file path " + path + " does
not exist.");
}
{code}
I suspect this has something to do with consistency guarantees provided by GCS.
I'm wondering if it's better to fail lazily at a later stage (e.g. when opening
the stream and it doesn't work).
After removing this check, everything works as expected. I can also run a batch
WordCount job against the same file.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
