Strategies for increasing availability with Connection Managers
You can increase the resiliency of your client/server communication environment.
Install multiple network-interface cards (NICs) on client, Connection Manager, and database-server hosts
You can prevent a network-connectivity failure by installing multiple NICs on each host in a connection-management domain. If a NIC fails, the host can use other available NICs.
Install multiple Connection Managers
You can prevent a Connection Manager from becoming a single point of failure in your system by installing multiple Connection Managers to manage a domain.
Install Connection Managers separate from database servers
To prevent simultaneous Connection Manager and database server failure if a host fails, install Connection Managers on hosts that are not running database servers.
Use keywords in Connection Manager service-level-agreements
If a Connection Manager manages a high-availability cluster, you can use the cluster keywords to maintain connection consistency after failover. If database servers in a cluster switch roles after failover, the Connection Manager can use cluster keywords to continue directing client connection requests to the appropriate cluster-server type.
- HDR - High-availability data replication server
- RSS - Remote standalone secondary server
- SDS - Shared-disk secondary server
- PRI - Primary server
- PRIMARY - Primary server
Use group entries in sqlhosts files and Connection Manager configuration files
Create database-server groups in the host sqlhosts file of each Connection Manager that manages a high-availability cluster. Then, specify sqlhosts groups, rather than individual server names, in Connection Manager configuration files. If you specify database server names for a connection unit's ONEDB_SERVER parameter, and those specified servers are offline when a Connection Manager restarts, the Connection Manager is unable to reconnect to the cluster. If you specify a sqlhosts group for a connection unit's ONEDB_SERVER parameter, and at least one of the group's database servers is online when a Connection Manager restarts, the Connection Manager can reconnect to the cluster.
c=1
option for sqlhosts group
entries so that the connection-attempt starting point for a list of
group members is random. For example:#dbservername nettype hostname servicename options
my_servers - - c=1
server_1 onsoctcp host_1 port_1 g=my_servers
server_2 onsoctcp host_2 port_2 g=my_servers
server_3 onsoctcp host_3 port_3 g=my_servers