I have a datastax cassandra cluster with 8 nodes. The keyspace used by the application contains about 400 Tables. The parameter write_request_timeout_in_ms in the cassandra.yaml is set to 2000ms (default).
The default value is high enough for most tables. However, I for only two tables I require a much higher write_request_timeout. I know that stuff such as bloomfilter false-positive chance or compaction strategy can be configured per table.
Is that possible to do the same for timeouts and if so then how?
Regards
It isn't possible to configure different write timeouts because all writes are persisted to the same commitlog disk.
A coordinator will return a write timeout if not enough replicas (based on the write consistency level) acknowledged the write (to the commitlog disk) because the disk is busy.
Since there is only one commitlog disk on each node, it makes no sense to have different write timeouts. This in fact raises another question -- what problem are you trying to solve?
Increasing timeouts is almost never the right thing to do since all it does is hide the problem. You need to identify the root of the issue and fix it. Cheers!