[
https://issues.apache.org/jira/browse/CASSANDRA-10794?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15042088#comment-15042088
]
Sam Tunnicliffe commented on CASSANDRA-10794:
---------------------------------------------
I think we probably can actually:
* Fix the table definition in {{AuthKeyspace}}, then when a node is restarted
it'll create the new table.
* Copy any existing data from the old table to the new in
{{CassandraAuthorizer::setup}}
* While the old table remains present in schema, double all writes to both the
old & new tables, but read only from the old one.
* Once all nodes are updated, drop the old table, reads and writes only hit the
new table.
Fortunately, that table is only on used during {{GRANT/REVOKE/DROP}} so doing
the check & double writes shouldn't be too much of a problem. There are some
edge cases to think through around performing {{GRANT/REVOKE/DROP}} on
non-upgraded nodes while the cluster is in a mixed state, but I think we should
be able to figure them out.
I can't believe I typo'd the name of a system table.
> System table name resource_role_permissons_index is spelt wrong!
> ----------------------------------------------------------------
>
> Key: CASSANDRA-10794
> URL: https://issues.apache.org/jira/browse/CASSANDRA-10794
> Project: Cassandra
> Issue Type: Bug
> Reporter: Max Bowsher
> Assignee: Sam Tunnicliffe
> Priority: Trivial
> Fix For: 3.0.x, 3.x
>
>
> System table name resource_role_permissons_index is spelt wrong!
> "permissons" is missing an "i"
> Fixing that isn't going to be fun, though :-(
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
