SASI is in early experiment and had many major problems. for example, "nodetool repair breaks SASI index"
https://issues.apache.org/jira/browse/CASSANDRA-13403 "OOM when using SASI index" https://issues.apache.org/jira/browse/CASSANDRA-12662 I would not use SASI index for production. Fay On Tue, Aug 15, 2017 at 7:03 AM, Vladimir Yudovin <vla...@winguzone.com> wrote: > Hi, > > I recently encountered with strange issue. > Assuming there is table > > id PRIMARY KEY > indexed text > column text > > CREATE custom index on table(indexed) using '...SASIIndex' > > I inserted row like *id=0, indexed='string1', column='just string'* > > When I did *SELECT * FROM table WHERE id=0 AND indexed='string1' *no rows > is returned, while *SELECT * FROM table WHERE id=0 AND column =' just > string' ALLOW FILTERING* returned the row. > > In reality this table consist about 70 columns, 20M rows and probably > total 30G on disk with RF=2 and three nodes. I guess this issue is somehow > linked to memory usage, as I run some different heavy queries, then cluster > become unresponsible to cqlsh and nodetool. I sow GC messages in gc.log, so > I restated cluster and all returned to normal. > > I would expect the whole query to fail with timeout or some error message, > but not to silently return zero rows. > Any thoughts? > > Best regards, Vladimir Yudovin, > *Winguzone <https://winguzone.com?from=list> - Cloud Cassandra Hosting* > >