The problem with row level ttls in this regard is that the scope of a
particular compaction may not include all versions of a given row. So just
because a primary key liveness denoting an expired row may be written to
the new SSTable, it doesn't necessarily mean that an index should purge its
entry/entries for that row. That said, it should probably be the
responsibility of the index implementation to manage that, but at the
moment the {{onPrimaryKeyLivenessInfo}} event during compaction doesn't
cause the registered indexes to be notified. I've opened
https://issues.apache.org/jira/browse/CASSANDRA-11329 for this.On Wed, Mar 9, 2016 at 4:05 PM, Eduardo Alonso <eduardoalo...@stratio.com> wrote: > Hi, > > We have been investigating how to include in our 2i implementation the > ability to index TTL expirable Cells in Cassandra 3.x. > > Reading comments in o.a.c.index.Index.Indexer.removeRows it seems that this > method is called when a compaction detects that a cell has expired. > > I dont know if this is correct, so the question is: > > How does the non-columnFamily based 2i get notified when a row ttl has > expired? > > I have checked the only two calls to Indexer.removeRows in > o.a.c.index.SecondaryIndexManager.IndexGCTransaction.commit in the > followings releases and it seems that has not changed at all. > > Sam Tunnicliffe, any ideas about this?? > > Regards > > > Eduardo Alonso > Vía de las dos Castillas, 33, Ática 4, 3ª Planta > 28224 Pozuelo de Alarcón, Madrid > Tel: +34 91 828 6473 // www.stratio.com // *@stratiobd > <https://twitter.com/StratioBD>* >
