As far as I remember, 2i index consistency is ensured and repaired at read time to improve performance and avoid the dreadful read-before-write: http://www.datastax.com/dev/blog/improving-secondary-index-write-performance-in-1-2
On Sun, Jun 22, 2014 at 5:24 PM, Jack Krupansky <j...@basetechnology.com> wrote: > The Cassandra 2.0 doc says that “Indexes are built in the background > automatically, without blocking reads or writes.” IOW, no guarantee that > the 2i will have been updated by the time you attempt a query on it. A GC > or just general busyness of the node could delay the 2i update, even though > it may generally complete quicker than you can do a query. > > See: > > http://www.datastax.com/documentation/cql/3.1/cql/ddl/ddl_build_index_c.html > > AFAICT, there is no mechanism for guaranteeing that a 2i has been updated. > > -- Jack Krupansky > > *From:* Tom van den Berge <t...@drillster.com> > *Sent:* Thursday, June 19, 2014 5:26 AM > *To:* user@cassandra.apache.org > *Subject:* Are writes to indexes performed asynchronously? > > Hi, > I have a column family with a secondary index on one of its columns. I > noticed that when I write a row to the column family, and immediately query > that row through the secondary index, every now and then it won't give any > results. > > Could it be that Cassandra performs the write to the internal index column > family asynchronously? That might explain this behaviour. > > In other words, when writing to a indexed column family, is there, or can > there be any guarantee that the write to index is completed when the write > to the original column family is completed? > > I'm using a single-node cluster, with consistency level ONE. > > Thanks, > Tom >
