This is a tough one. One thing I can think of is to use Spark/Spark SQL to run ad-hoc queries on C* cluster. You can post on "Spark Cassandra Connector" user group.
On Tue, Mar 3, 2015 at 10:18 AM, Yulian Oifa <oifa.yul...@gmail.com> wrote: > Hello > Initially problem is that customer wants to have an option for ANY query , > which does not fits good with NOSQL.However the size of data is too big for > Relational DB. > There are no typical queries on the data, there are 10 fields , based on > which ( any mix of them also ) queries should be made. > Till now i allowed only single field query ( in specific cases with 2 > fields ) so i had indexes CFs for each field and that solved a > problem.Since now i need compounds , sometimes 5-6 fields i either need to > iterate over index CF based on some column ( or may be read several indexes > and find common ids ) or create some index that will allow me to read data > based on any part. Creating index for each group of fields of course is not > an option since number of indexes will be huge , and disk usage will be too > big. > > Best regards > Yulian Oifa > > On Mon, Mar 2, 2015 at 5:33 PM, Kai Wang <dep...@gmail.com> wrote: > >> AFIK it's not possible. The fact you need to query the data by partial >> row key indicates your data model isn't proper. What are your typical >> queries on the data? >> >> On Sun, Mar 1, 2015 at 7:24 AM, Yulian Oifa <oifa.yul...@gmail.com> >> wrote: >> >>> Hello to all. >>> Lets assume a scenario where key is compound type with 3 types in it ( >>> Long , UTF8, UTF8 ). >>> Each row stores timeuuids as column names and empty values. >>> Is it possible to retreive data by single key part ( for example by long >>> only ) by using java thrift? >>> >>> Best regards >>> Yulian Oifa >>> >>> >>> >> >
