[
https://issues.apache.org/jira/browse/HIVE-21731?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
mahesh kumar behera updated HIVE-21731:
---------------------------------------
Attachment: HIVE-21731.02.patch
> Hive import fails, post upgrade of source 3.0 cluster, to a target 4.0
> cluster with strict managed table set to true.
> ---------------------------------------------------------------------------------------------------------------------
>
> Key: HIVE-21731
> URL: https://issues.apache.org/jira/browse/HIVE-21731
> Project: Hive
> Issue Type: Bug
> Reporter: mahesh kumar behera
> Assignee: mahesh kumar behera
> Priority: Major
> Labels: pull-request-available
> Attachments: HIVE-21731.01.patch, HIVE-21731.02.patch
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> The scenario is
> # Replication policy is set with hive 3.0 source cluster (strict managed
> table set to false) and hive 4.0 target cluster with strict managed table set
> true.
> # User upgrades the 3.0 source cluster to 4.0 cluster using upgrade tool.
> # The upgrade converts all managed tables to acid tables.
> # In the next repl dump, user sets hive .repl .dump .include .acid .tables
> and hive .repl .bootstrap. acid. tables set true triggering bootstrap of
> newly converted ACID tables.
> # As the old tables are non-txn tables, dump is not filtering the events
> even tough bootstrap acid table is set to true. This is causing the repl load
> to fail as the write id is not set in the table object.
> # If we ignore the event replay, the bootstrap is failing with dump
> directory mismatch error.
> The fix should be
> # Ignore dumping the alter table event if bootstrap acid table is set true
> and the alter is converting a non-acid table to acid table.
> # In case of bootstrap during incremental load, ignore the dump directory
> property set in table object.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
