[
https://issues.apache.org/jira/browse/CASSANDRA-4670?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roland Gude updated CASSANDRA-4670:
-----------------------------------
Attachment: compaction2.log
compaction2.log contains a little bit more info. still limited to lines about
the relevant index - anything that was in the logs near by the compaction stuff.
> LeveledCompaction destroys secondary indexes
> --------------------------------------------
>
> Key: CASSANDRA-4670
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4670
> Project: Cassandra
> Issue Type: Bug
> Affects Versions: 1.1.4, 1.1.5
> Reporter: Roland Gude
> Attachments: compaction2.log, compaction.log
>
>
> When LeveledCompactionStrategy is active on a ColumnFamily with an Index
> enabled on TTL Columns, the Index is not working correctly, because the
> compaction is throwing away index data very aggressively.
> Steps to reproduce:
> create a cluster with a columnfamily with an indexed column and leveled
> compaction:
> create column family CorruptIndex
> with column_type = 'Standard'
> and comparator = 'TimeUUIDType'
> and default_validation_class = 'BytesType'
> and key_validation_class = 'BytesType'
> and read_repair_chance = 0.5
> and dclocal_read_repair_chance = 0.0
> and gc_grace = 864000
> and min_compaction_threshold = 4
> and max_compaction_threshold = 32
> and replicate_on_write = true
> and compaction_strategy =
> 'org.apache.cassandra.db.compaction.LeveledCompactionStrategy'
> and caching = 'NONE'
> and column_metadata = [
> {column_name : '00000003-0000-1000-0000-000000000000',
> validation_class : BytesType,
> index_name : 'idx_corrupt',
> index_type : 0}];
> in that cf insert expiring data (expiration date should be in the far future
> for the sake of this test)
> query the data by index:
> get CorruptIndex where 00000003-0000-1000-0000-000000000000=utf8(‘value’)
> see results (should be correct for some time)
> wait for leveled compaction to compact the index
> query the data by index:
> get CorruptIndex where 00000003-0000-1000-0000-000000000000=utf8(‘value’)
> see results (are empty)
> trigger rebuild index via nodetool
> query the data by index:
> get CorruptIndex where 00000003-0000-1000-0000-000000000000=utf8(‘value’)
> should be corretc again
> wait for leveled compaction to compact the index
> query the data by index:
> get CorruptIndex where 00000003-0000-1000-0000-000000000000=utf8(‘value’)
> see results (are empty)
> repeat until bored
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
