Hi
My NdbCluster deployment is as follows:
Server 1 - ndbd + mysqld
Server 2 - ndbd + mysqld
Server 3 - ndb_mgm
NoOfReplicas = 2, so each of servers 1/2 hold full data-set
As I can see in tcpdump mysqld processes on both servers connect and fetch data both from local and remote ndbd processes.
I think that by avoiding network overhead I can get better performance by forcing mysqld process to query ndbd co-located on the same server, rather than remote one.
Is there any possibility to configure mysqld on server 1 to prefer fetching data from ndbd on server 1 and use data from server 2 only in case of failure of local ndbd?
Maybe some sort of cost-based query routing used in WAN environments can be used here?
Any hints?
My NdbCluster deployment is as follows:
Server 1 - ndbd + mysqld
Server 2 - ndbd + mysqld
Server 3 - ndb_mgm
NoOfReplicas = 2, so each of servers 1/2 hold full data-set
As I can see in tcpdump mysqld processes on both servers connect and fetch data both from local and remote ndbd processes.
I think that by avoiding network overhead I can get better performance by forcing mysqld process to query ndbd co-located on the same server, rather than remote one.
Is there any possibility to configure mysqld on server 1 to prefer fetching data from ndbd on server 1 and use data from server 2 only in case of failure of local ndbd?
Maybe some sort of cost-based query routing used in WAN environments can be used here?
Any hints?