[
https://issues.apache.org/jira/browse/CASSANDRA-8865?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Constance Eustace updated CASSANDRA-8865:
-----------------------------------------
Description:
We are upgrading our dev cluster. 2.0.10 to 2.1.3
The indexes are behaving very strangely.
create index definition_bundle__BundleDefSkuIDXTest on
definition_bundle.entity_bundledef(e_entlinks) ;
definition_bundle> select column_name, index_name, index_options, index_type,
component_index from system.schema_columns where keyspace_name =
'definition_bundle' and columnfamily_name = 'entity_bundledef';
column_name | index_name | index_options |
index_type | component_index
-------------+----------------------------------------+---------------+------------+-----------------
bundle_sku | definition_bundle__BundleDefSkuIDX | {} |
COMPOSITES | 1
e_entid | null | null |
null | null
e_entlinks | definition_bundle__bundledefskuidxtest | {} |
COMPOSITES | 1
NOTICE THE AUTO-DOWNCASE of our newly created index. The index that already
existed is NOT AUTO-DOWNCASED. I don't know if this is recent or not.
We cannot drop the mixed case index. Nodetool index reconstruction did not
work. Indexes are doing very weird things.
Hm. UPDATE:
This did successfully delete the index:
drop index "definition_bundle__BundleDefSkuIDX";
Anyway, it looks like there is some upcase/downcase assumptions not being
properly done.... somewhere, either in upgrades or similar stuff.
We will probably drop our indexes and recreate them.
was:
We are upgrading our dev cluster.
The indexes are behaving very strangely.
create index definition_bundle__BundleDefSkuIDXTest on
definition_bundle.entity_bundledef(e_entlinks) ;
definition_bundle> select column_name, index_name, index_options, index_type,
component_index from system.schema_columns where keyspace_name =
'definition_bundle' and columnfamily_name = 'entity_bundledef';
column_name | index_name | index_options |
index_type | component_index
-------------+----------------------------------------+---------------+------------+-----------------
bundle_sku | definition_bundle__BundleDefSkuIDX | {} |
COMPOSITES | 1
e_entid | null | null |
null | null
e_entlinks | definition_bundle__bundledefskuidxtest | {} |
COMPOSITES | 1
NOTICE THE AUTO-DOWNCASE of our newly created index. The index that already
existed is NOT AUTO-DOWNCASED. I don't know if this is recent or not.
We cannot drop the mixed case index. Nodetool index reconstruction did not
work. Indexes are doing very weird things.
Hm. UPDATE:
This did successfully delete the index:
drop index "definition_bundle__BundleDefSkuIDX";
Anyway, it looks like there is some upcase/downcase assumptions not being
properly done.... somewhere, either in upgrades or similar stuff.
We will probably drop our indexes and recreate them.
> DROP INDEX name case sensitivity causing errors in cass upgrade 2.0.10 to
> 2.1.3
> -------------------------------------------------------------------------------
>
> Key: CASSANDRA-8865
> URL: https://issues.apache.org/jira/browse/CASSANDRA-8865
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Environment: Amazon, single node, ubuntu 14.04, jdk 7
> Reporter: Constance Eustace
>
> We are upgrading our dev cluster. 2.0.10 to 2.1.3
> The indexes are behaving very strangely.
> create index definition_bundle__BundleDefSkuIDXTest on
> definition_bundle.entity_bundledef(e_entlinks) ;
> definition_bundle> select column_name, index_name, index_options, index_type,
> component_index from system.schema_columns where keyspace_name =
> 'definition_bundle' and columnfamily_name = 'entity_bundledef';
> column_name | index_name | index_options |
> index_type | component_index
> -------------+----------------------------------------+---------------+------------+-----------------
> bundle_sku | definition_bundle__BundleDefSkuIDX | {} |
> COMPOSITES | 1
> e_entid | null | null |
> null | null
> e_entlinks | definition_bundle__bundledefskuidxtest | {} |
> COMPOSITES | 1
> NOTICE THE AUTO-DOWNCASE of our newly created index. The index that already
> existed is NOT AUTO-DOWNCASED. I don't know if this is recent or not.
> We cannot drop the mixed case index. Nodetool index reconstruction did not
> work. Indexes are doing very weird things.
> Hm. UPDATE:
> This did successfully delete the index:
> drop index "definition_bundle__BundleDefSkuIDX";
> Anyway, it looks like there is some upcase/downcase assumptions not being
> properly done.... somewhere, either in upgrades or similar stuff.
> We will probably drop our indexes and recreate them.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
