Skip to content

Releases: linkedin/iceberg

0.0.13

07 Feb 17:36
Compare
Choose a tag to compare
  • Incremental api
  • Upstream changes
  • Bug fixes

Publish upstream patches

03 Dec 06:17
Compare
Choose a tag to compare
Pre-release
0.0.12

Disable test broken due to LIHADOOP-49587

Add a new Iceberg-runtime module

04 Nov 22:32
Compare
Choose a tag to compare
Pre-release
0.0.11

Add a iceberg-runtime shaded module (#12)

Pull new changes from upstream

03 Nov 07:35
Compare
Choose a tag to compare
Pre-release
Add custom hive catalog to not override existing Hive metadata (#10)

Add custom hive catalog to not override existing Hive metadata

0.0.9

25 Oct 03:54
Compare
Choose a tag to compare
0.0.9 Pre-release
Pre-release

Added custom Hive catalog

0.0.8

23 Oct 17:26
Compare
Choose a tag to compare
0.0.8 Pre-release
Pre-release
  • Downgrade jackson to get around security issue
  • Also fix commit hashes so that they are in sync with upstream Iceberg

Release changes for automatic schema evolution

23 Oct 16:31
Compare
Choose a tag to compare
0.0.7

LIHADOOP-49587: [Writer schema evolution] Update ReassignIds to assig…

0.0.6: UpdateSchema api for addition, updates and deletes (#4)

21 Oct 23:24
1f82c94
Compare
Choose a tag to compare
* UpdateSchema api for addition, updates and deletes

* UpdateSchema api for addition, updates and deletes: Address comments

Releasing 0.0.5

21 Oct 23:18
1f82c94
Compare
Choose a tag to compare
Releasing 0.0.5 Pre-release
Pre-release
UpdateSchema api for addition, updates and deletes (#4)

* UpdateSchema api for addition, updates and deletes

* UpdateSchema api for addition, updates and deletes: Address comments

Incorporate latest changes

15 Oct 00:03
Compare
Choose a tag to compare
Pre-release
Fix connection leak in Hive table tests (#538)

This is another attempt to fix the connection leak in the Hive table tests. The current theory is that the cache in HiveCatalogs is not behaving correctly. Because it uses weak keys, garbage collection will almost always remove the cached catalog between SQL queries, and the next query will result in a new connection to the metastore. The garbage collected catalog is not closed in this case.

This fixes the leak in two ways:

1. Use a fixed duration after the last access, 10 minutes, to expire catalogs
2. Because catalogs are expired instead of garbage collected, use a removal listener to close
3. Add a catch-all finalizer to HiveCatalog to ensure that catalogs are closed when leaked, and a stack trace is logged to help fix the leak