[ https://issues.apache.org/jira/browse/CASSANDRA-5411?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13624102#comment-13624102 ]
Arya Goudarzi commented on CASSANDRA-5411: ------------------------------------------ The notion of delivering or truncating hints before changing gc_grace isn't documented anywhere. IMHO it is not fair to the operator to have to remember and additional step in a scenario that can happen to anyone; specially, when several people in chatroom are dealing with tombstone cleanups situations and changing gc_grace has become part of their routine. So, I suggest re-opening it, until there is a feasible solution. I wished if I had the capacity to help you with this myself, but frankly I am running a one-man operation here and am kinda tied up. > Unexpired hints can potentially create previously deleted rows > -------------------------------------------------------------- > > Key: CASSANDRA-5411 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5411 > Project: Cassandra > Issue Type: Bug > Components: Core > Affects Versions: 1.1.10 > Environment: Ubuntu 10.04 > Reporter: Arya Goudarzi > Assignee: Brandon Williams > > Per discussion here: > http://www.mail-archive.com/user@cassandra.apache.org/msg28905.html > Hints are TTLed to lowest gc_grace_seconds. If we have hints that are not > replayed, and the administrator updates schema to have lower gc_grace seconds > on some CF, chances are compactions will delete tombstones and history of > delete gets lost. If the deleted row is in some hints CF, since the TTL is > not updated, chances are it gets replayed and the deleted row comes back to > life. > I think it would be nice as part of schema migration, to also update the TTLs > on hints that already exists on any nodes hints column family. -- 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