Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
2.7.4
-
None
-
None
Description
Discussion around SPARK-22587 highlights how per-fs notions of a canonical URI make it hard to determine if a file is on a specific filesystem, or, put differently, if two filesystems are equivalent.
You can't reliably use this.getUri == that.getUri as it doesn't handle FQDN == unqualified DN, bit you can't do nslookup as HDFS HA doesn't use hosnames.
If FileSystem.getCanonicalUri() were public, then this could be used to compare things consistently.
needs: filesystem.md coverage; contract test (two filesystem instances are equal, different filesystems aren't). Or at least: this method never returns null.
Attachments
Issue Links
- relates to
-
SPARK-22587 Spark job fails if fs.defaultFS and application jar are different url
- Resolved