[
https://issues.apache.org/jira/browse/CASSANDRA-17503?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Paul Ayers updated CASSANDRA-17503:
-----------------------------------
Description:
Testing rebuild in C* 4.0.1, I ran into a scenario where 'nodetool rebuild' was
ran without any options in a new DC and the system.available_ranges_v2 table
was populated such that every subsequent run of 'nodetool rebuild <sourceDC>'
streamed no data.
The problem was that the first run looked at only the new DC so no data was
streamed, then the next run that provided the correct source DC still resulted
in no streamed data.
Once the system.available_ranges_v2 table was truncated, rebuild worked as
expected.
The request here is to include the same code that allows the following options
to be used in DSE so one could easily resolve this by using '-m reset' in a
similar scenario:
-m, --mode mode
* normal - conventional behavior, streams only ranges that are not already
locally available
* refetch - resets locally available ranges, streams all ranges but leaves
current data untouched
* reset - resets the locally available ranges, removes all locally present
data (like a TRUNCATE), streams all ranges
* reset-no-snapshot - (like reset) resets the locally available ranges,
removes all locally present data (like a TRUNCATE), streams all ranges but
prevents a snapshot even if auto_snapshot is enabled
Protecting against updating the available_ranges table if a local DC was chosen
might also be a good idea.
was:
Testing rebuild in C* 4.0.1, I ran into a scenario where 'nodetool rebuild' was
ran without any options in a new DC and the system.available_ranges_v2 table
was populated such that every subsequent run of 'nodetool rebuild <sourceDC>'
streamed no data.
The problem was that the first run looked at only the new DC so no data was
streamed, then the next run that provided the correct source DC still resulted
in no streamed data.
Once the system.available_ranges_v2 table was truncated, rebuild worked as
expected.
The request here is to include the same code that allows the following options
to be used in DSE so one could easily resolve this by using '-m reset' in a
similar scenario:
-m, --mode mode
* normal - conventional behavior, streams only ranges that are not already
locally available
* refetch - resets locally available ranges, streams all ranges but leaves
current data untouched
* reset - resets the locally available ranges, removes all locally present
data (like a TRUNCATE), streams all ranges
* reset-no-snapshot - (like reset) resets the locally available ranges,
removes all locally present data (like a TRUNCATE), streams all ranges but
prevents a snapshot even if auto_snapshot is enabled
> Request to port the DSE rebuild options to Cassandra
> ----------------------------------------------------
>
> Key: CASSANDRA-17503
> URL: https://issues.apache.org/jira/browse/CASSANDRA-17503
> Project: Cassandra
> Issue Type: New Feature
> Reporter: Paul Ayers
> Priority: Normal
>
> Testing rebuild in C* 4.0.1, I ran into a scenario where 'nodetool rebuild'
> was ran without any options in a new DC and the system.available_ranges_v2
> table was populated such that every subsequent run of 'nodetool rebuild
> <sourceDC>' streamed no data.
> The problem was that the first run looked at only the new DC so no data was
> streamed, then the next run that provided the correct source DC still
> resulted in no streamed data.
> Once the system.available_ranges_v2 table was truncated, rebuild worked as
> expected.
> The request here is to include the same code that allows the following
> options to be used in DSE so one could easily resolve this by using '-m
> reset' in a similar scenario:
> -m, --mode mode
> * normal - conventional behavior, streams only ranges that are not already
> locally available
> * refetch - resets locally available ranges, streams all ranges but leaves
> current data untouched
> * reset - resets the locally available ranges, removes all locally present
> data (like a TRUNCATE), streams all ranges
> * reset-no-snapshot - (like reset) resets the locally available ranges,
> removes all locally present data (like a TRUNCATE), streams all ranges but
> prevents a snapshot even if auto_snapshot is enabled
> Protecting against updating the available_ranges table if a local DC was
> chosen might also be a good idea.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
