In Cassandra 1.1 the schema is no longer a full migration history. Before that each schema change is recorded in the table and they all have to be replayed when a node bootstraps. Also 1.1 has some bug ATM that means you should not switch to it.
People tend to say things like "with cassandra X you now can have more CF and KS's" but back in the day the thinking was "You only need 1" Hector has a concept of "virtual keyspaces" which is a model that is likely to give you more success then 1500 CFs. I am not a fan of dynamically creating and tearing down CFS on the fly or 1 per customer designs. When the wrinkles come out of 1.1 and up upgrade the schema size should shrink down. On Fri, May 25, 2012 at 6:25 AM, Sasha Yanushkevich <yanus...@gmail.com> wrote: > Hi All, > > I have a question about schema cleaning in cassandra. > > I use cassandra version 1.0.9. I have 5 keyspaces and about 1500 column > family per keyspace. After dynamically creating and deleting CF my schema's > sstables size were very high. For example size of Migrations was 45 GB and > Schema sstable size was 45 GB whereas all data size was about 10 GB. > > Why schema grow was so fast? And how may I cleanup schema or later schema > cleanup itself? > > > -- > Best regards, > Alexander >
