On Fri, Dec 4, 2015 at 11:44 AM, Anuj Wadehra <anujw_2...@yahoo.co.in> wrote:
> Did u say "longer than gc_grace_seconds" ? > Wont deletes pop back during repair? > Unfortunately, you are correct. Since CASSANDRA-4905 [1], such tombstones will not be propagated. The actual way to fully repair a node that has been down greater than gc_grace_seconds, without rebootstrapping it and decreasing unique replica count, is to : 1) extend gc_grace_seconds to cover all un-purged tombstones 2) repair 3) change gc_grace_seconds back =Rob [1] thanks to driftx @ IRC for clarification
