[
https://issues.apache.org/jira/browse/CASSANDRA-10798?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
MASSIMO CELLI updated CASSANDRA-10798:
--------------------------------------
Description:
on C* 2.1.12 if you create a new table and a directory with the same name
already exist under the keyspace then C* will simply use that directory rather
than creating a new one that has uuid in the name.
Even if you drop and recreate the same table it will still use the previous dir
and never switch to a new one with uuid. This can happen on one of the nodes in
the cluster while the other nodes will use the uuid format for the same table.
For example I dropped and recreated the same table three times in this test on
a two nodes cluster
node1
drwxr-xr-x 3 cassandra cassandra 4096 Dec 1 23:47 mytable
node2
drwxr-xr-x 3 cassandra cassandra 4096 Dec 1 23:47
mytable-678a7e31988511e58ce7cfa0aa9730a2
drwxr-xr-x 3 cassandra cassandra 4096 Dec 1 23:41
mytable-cade4ee1988411e58ce7cfa0aa9730a2
drwxr-xr-x 2 cassandra cassandra 4096 Dec 1 23:47
mytable-db1b9b41988511e58ce7cfa0aa9730a2
This seems to break the changes introduced by CASSANDRA-5202
was:
on C* 2.1.12 if you create a new table and a directory with the same name
already exist under the keyspace then C* will simply use that directory rather
than creating a new one that has uuid in the name.
Even if you drop and recreate the same table it will still use the previous dir
and never switch to a new one with uuid. This can happen on one of the nodes in
the cluster while the other nodes will use the uuid format for the same table.
This seems to break the changes introduced by CASSANDRA-5202
> C* 2.1 doesn't create dir name with uuid if dir is already present
> ------------------------------------------------------------------
>
> Key: CASSANDRA-10798
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10798
> Project: Cassandra
> Issue Type: Bug
> Components: Local Write-Read Paths
> Environment: C* 2.1.11
> Reporter: MASSIMO CELLI
> Fix For: 2.1.x
>
>
> on C* 2.1.12 if you create a new table and a directory with the same name
> already exist under the keyspace then C* will simply use that directory
> rather than creating a new one that has uuid in the name.
> Even if you drop and recreate the same table it will still use the previous
> dir and never switch to a new one with uuid. This can happen on one of the
> nodes in the cluster while the other nodes will use the uuid format for the
> same table.
> For example I dropped and recreated the same table three times in this test
> on a two nodes cluster
> node1
> drwxr-xr-x 3 cassandra cassandra 4096 Dec 1 23:47 mytable
> node2
> drwxr-xr-x 3 cassandra cassandra 4096 Dec 1 23:47
> mytable-678a7e31988511e58ce7cfa0aa9730a2
> drwxr-xr-x 3 cassandra cassandra 4096 Dec 1 23:41
> mytable-cade4ee1988411e58ce7cfa0aa9730a2
> drwxr-xr-x 2 cassandra cassandra 4096 Dec 1 23:47
> mytable-db1b9b41988511e58ce7cfa0aa9730a2
> This seems to break the changes introduced by CASSANDRA-5202
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
