[
https://issues.apache.org/jira/browse/HIVE-19219?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sankar Hariappan updated HIVE-19219:
------------------------------------
Description:
This is the case where the events were deleted on source because of old event
purging and hence min(source event id) > target event id (last replicated event
id).
Repl dump should fail in this case so that user can drop the database and
bootstrap again.
Cleaner thread is concurrently removing the expired events from
NOTIFICATION_LOG table. So, it is necessary to check if the current dump missed
any event while dumping. Also, it is guaranteed that cleaner would clean
sequentially based on event id. So, it is enough to check if the first event >
last event id is still there in NOTIFICATION_LOG table even after dumping all
events.
If it is not found, then likely some events missed in the dump and hence throw
error.
was:
This is the case where the events were deleted on source because of old event
purging and hence min(source event id) > target event id (last replicated event
id).
Repl dump should fail in this case so that user can drop the database and
bootstrap again.
The next incremental repl dump could check if the events fetched from source
notification_log table is retrieved in continuous sequence with no events
missing. If any event is missing, it should throw error.
> Incremental REPL DUMP should throw error if requested events are cleaned-up.
> ----------------------------------------------------------------------------
>
> Key: HIVE-19219
> URL: https://issues.apache.org/jira/browse/HIVE-19219
> Project: Hive
> Issue Type: Bug
> Affects Versions: 3.0.0
> Reporter: Sankar Hariappan
> Assignee: Sankar Hariappan
> Priority: Major
> Labels: DR, pull-request-available, replication
> Fix For: 3.1.0
>
> Attachments: HIVE-19219.01.patch
>
>
> This is the case where the events were deleted on source because of old event
> purging and hence min(source event id) > target event id (last replicated
> event id).
> Repl dump should fail in this case so that user can drop the database and
> bootstrap again.
> Cleaner thread is concurrently removing the expired events from
> NOTIFICATION_LOG table. So, it is necessary to check if the current dump
> missed any event while dumping. Also, it is guaranteed that cleaner would
> clean sequentially based on event id. So, it is enough to check if the first
> event > last event id is still there in NOTIFICATION_LOG table even after
> dumping all events.
> If it is not found, then likely some events missed in the dump and hence
> throw error.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
