You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Michael Semb Wever (Jira)" <ji...@apache.org> on 2025/12/01 20:47:00 UTC

[jira] [Updated] (CASSANDRA-21050) Missing rows on read after schema recreation with dropped frozen collection/tuple

     [ https://issues.apache.org/jira/browse/CASSANDRA-21050?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Michael Semb Wever updated CASSANDRA-21050:
-------------------------------------------
    Description: 
See `MissingRowsOnDroppedColumnSchemaRecreateTest.testMissingRows()`
in https://github.com/apache/cassandra/compare/cassandra-5.0...thelastpickle:cassandra:mck/21050/5.0 

Is failing because not all rows are being returned from the tables with dropped frozen collection/tuple after the schema recreation.


Observations are:
- Only happens in CC5 (but also with test data from Cassandra 4.1 and 5.0)
- Only needs data in sstables (i.e. the commitlogs can flushed and ignored)
- Does not happen if the *-Statistics.db files are deleted
- More than half the rows are missing, there's not yet any discerning pattern to which rows are missing, and which rows are missing changes test run to test run...
- There is only one sstable
- The column types are the same before and after schema recreation.
- DeserialisationHeader.isDropped(cell) returns the same on failing and passing tests.
- Everything in system_schema tables is the same after schema recreation as it was before.

Is this a runtime or jvm-dtest bug ? 
I believe it is n 4.1 as well.

  was:
See `MissingRowsOnDroppedColumnSchemaRecreateTest.testMissingRows()`
in https://github.com/apache/cassandra/compare/cassandra-5.0...thelastpickle:cassandra:mck/21050/5.0 

Is failing because not all rows are being returned from the tables with dropped frozen collection/tuple after the schema recreation.


Observations are:
- Only happens in CC5 (but also with test data from Cassandra 4.1 and 5.0)
- Only needs data in sstables (i.e. the commitlogs can flushed and ignored)
- Does not happen if the *-Statistics.db files are deleted
- More than half the rows are missing, there's not yet any discerning pattern to which rows are missing, and which rows are missing changes test run to test run...
- There is only one sstable
- The column types are the same before and after schema recreation.
- DeserialisationHeader.isDropped(cell) returns the same on failing and passing tests.
- Everything in system_schema tables is the same after schema recreation as it was before.

Is this a runtime or jvm-dtest bug ? 


> Missing rows on read after schema recreation with dropped frozen collection/tuple
> ---------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-21050
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-21050
>             Project: Apache Cassandra
>          Issue Type: Bug
>          Components: Local/SSTable
>            Reporter: Michael Semb Wever
>            Priority: Normal
>             Fix For: 5.0.x, 5.x
>
>
> See `MissingRowsOnDroppedColumnSchemaRecreateTest.testMissingRows()`
> in https://github.com/apache/cassandra/compare/cassandra-5.0...thelastpickle:cassandra:mck/21050/5.0 
> Is failing because not all rows are being returned from the tables with dropped frozen collection/tuple after the schema recreation.
> Observations are:
> - Only happens in CC5 (but also with test data from Cassandra 4.1 and 5.0)
> - Only needs data in sstables (i.e. the commitlogs can flushed and ignored)
> - Does not happen if the *-Statistics.db files are deleted
> - More than half the rows are missing, there's not yet any discerning pattern to which rows are missing, and which rows are missing changes test run to test run...
> - There is only one sstable
> - The column types are the same before and after schema recreation.
> - DeserialisationHeader.isDropped(cell) returns the same on failing and passing tests.
> - Everything in system_schema tables is the same after schema recreation as it was before.
> Is this a runtime or jvm-dtest bug ? 
> I believe it is n 4.1 as well.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@cassandra.apache.org
For additional commands, e-mail: commits-help@cassandra.apache.org