How to override default_fetch_size for any query to a cassandra cluster or keyspace? - cassandra

Question

I have a cassandra 2.0.5 cluster setup with 3 nodes. Multiple services use the same cluster with separate keyspace. Due to the large size of the blob entry in the table, a query which goes over all the rows causes an OutOfMemory error and crashes the cluster. This is unacceptable for me as different services use the same cluster and one should not affect others.
Now, there is a way to restrict the number of rows which are loaded in memory at a time per query using the fetchSize parameter with a query with most of the drivers supporting cluster 2.0 protocol.
The default_fetch-size value is 5000, and I want to override this value to something smaller like 500 to avoid OOM error. I cannot assume all clients will use a small fetchSize while issuing an expensive query. Is there a way to do so? I cannot find any such configuration in cassandra.yaml.
Can I set this per keyspace level, so all queries to a particular keyspace will have a smaller default_fetch_size?

Answer 1

To set fetchsize on all statements, you can do that in QueryOptions when you build your session. Something like this,
Session session = Cluster.builder()
.addContactPoint("127.0.0.1")
.withQueryOptions(new QueryOptions().setFetchSize(100))
.build().connect();
If you want to have different fetchsize for different keyspaces then maintain session objects for each keyspace. Something like this,
Session sessionForKeyspace1 = Cluster.builder()
.addContactPoint("127.0.0.1")
.withQueryOptions(new QueryOptions().setFetchSize(100))
.build().connect("keyspace1");
Session sessionForKeyspace2 = Cluster.builder()
.addContactPoint("127.0.0.1")
.withQueryOptions(new QueryOptions().setFetchSize(200))
.build().connect("keyspace2");

Answer 2

What I know is you can only set fetch size of a particular statement. statement.setFetchSize(100);