Updating a dataset should not disable and re-enable explore

Description

Updating a dataset currently disables Explore, then calls upgrade() on the dataset, then enables Explore again.

Disabling, however, drops the existing Hive table, and that loses all the existing partitions. Enabling explore does not add these partitions back, it only creates an empty Hive table.

That means that every time the properties for a partitioned file set are updated, it loses all its partitions in Hive forever.

The same happens when creating a dataset and dataset.unchecked.upgrade is true in cdap-site.xml, and that means, with this flag on, redeploying an app deletes all partitions in the datasets created by this app.

Release Notes

Fixed an issue where updating the properties of a dataset deleted all of its partitions in Hive.

Activity

Show:

Ali AnwarAugust 12, 2016 at 8:39 PM

's change alter the Hive table instead of drop and create when updating a file set dataset:
https://github.com/caskdata/cdap/pull/6144

Fixed
Pinned fields
Click on the next to a field label to start pinning.

Details

Assignee

Reporter

Affects versions

Components

Fix versions

Priority

Created July 8, 2016 at 2:32 PM
Updated August 19, 2016 at 8:00 AM
Resolved July 27, 2016 at 6:47 AM