The fix for range tombstone performance [1] introduced a regression [2] that breaks slice queries against 1.1-format sstables. Running upgradesstables is a workaround, but this means that you can't do zero-downtime upgrades from 1.1 at CL.ONE.
I think we should withdraw 1.2.7 from the download mirrors and roll 1.2.7.1 with the fix. Thoughts? [1] https://issues.apache.org/jira/browse/CASSANDRA-5677 [2] https://issues.apache.org/jira/browse/CASSANDRA-5814 -- Jonathan Ellis Project Chair, Apache Cassandra co-founder, http://www.datastax.com @spyced
