Example of configuring connection management for a high-availability replication system

You can use a Connection Manager to route client connections for the participants of a replicate set and to control failover for high-availability clusters that participate in Enterprise Replication.

About this task

For this example, you have a grid that consists of four nodes. One of the nodes is a primary server in a high-availability cluster that consists of a primary server, an SD secondary server, an HDR secondary server, and an RS secondary server:
  • server_1a - ER Node 1, primary server
  • server_1b - SD secondary server
  • server_1c - HDR secondary server
  • server_1d - RS secondary server
  • server_2 - ER Node 2
  • server_3 - ER Node 3
  • server_4 - ER Node 4

The grid supports reporting services, which can run on any of the ER nodes.

Your system has the following needs
  • Client requests are directed to the ER node with the lowest transaction latency.
  • The system can withstand the failure of a Connection Manager.
  • The system can withstand a network-interface card (NIC) failure on each host.
  • The Connection Managers control failover for the cluster.
  • If failover occurs, it the SD secondary server takes priority over the HDR secondary server. The HDR secondary server takes priority over the RS secondary server.
  • If the primary server fails, the Connection Managers can still connect to the cluster after restarting.

Procedure

To configure connection management:
  1. Install at least two network interface cards on each host.
    This prevents the failure of a network interface card from causing Connection Manager or database server failure.
  2. Install two Connection Managers. Install each Connection Manager onto a different host, and do not install the Connection Managers onto the hosts that database servers are installed on.
    This installation strategy prevents a Connection Manager from becoming a single point of failure, and prevents the simultaneous failure of database servers and Connection Managers if a host fails.
  3. On each Connection Manager host, set the ONEDB_HOME environment to the directory the Connection Manager was installed into.
    Run the following command:
    setenv ONEDB_HOME path
  4. Create a configuration file in each Connection Manager installation's $ONEDB_HOME/etc directory.
    The first Connection Manager's configuration file is named cm_1.cfg and has the following entries:
    NAME connection_manger_1
    LOG 1
    LOGFILE $ONEDB_HOME/tmp/my_cm1_log.log
    LOCAL_IP 192.0.0.2,192.0.2.1
    
    REPLSET replicate_set_1
    {
    	ONEDB_SERVER	g_server_1,g_server_2,g_server_3,g_server_4
    	SLA report_1	DBSERVERS=ANY \
                 POLICY=LATENCY
    }
    
    CLUSTER cluster_1
    {
       ONEDB_SERVER g_server_1
       FOC ORDER=ENABLED \
           PRIORITY=1
       CMALARMPROGRAM $ONEDB_HOME/etc/CMALARMPROGRAM.sh
    }
    The second Connection Manager's configuration file is named cm_2.cfg and has the following entries:
    NAME connection_manger_2
    LOG 1
    LOGFILE $ONEDB_HOME/tmp/my_cm2_log_.log
    LOCAL_IP 192.0.2.2,192.0.2.3
    
    REPLSET replicate_set_1
    {
    	ONEDB_SERVER	g_server_1,g_server_2,g_server_3,g_server_4
    	SLA report_2	DBSERVERS=ANY \
                 POLICY=LATENCY
    }
    CLUSTER cluster_1
    {
       ONEDB_SERVER g_server_1
       FOC ORDER=ENABLED \
           PRIORITY=2
       CMALARMPROGRAM $ONEDB_HOME/etc/CMALARMPROGRAM.sh
    }
    The configuration file specifies the following information and behavior:
    • Logging is enabled, and the log files are $ONEDB_HOME/tmp/my_cm1_log.log and $ONEDB_HOME/tmp/my_cm2_log.log.
    • connection_manager_1 monitors 192.0.2.0 and 192.0.2.1 and connection_manager_2 monitors 192.0.2.2 and 192.0.2.3 for network failure.
    • When the Connection Managers start, they each search their sqlhosts files for g_server_1, g_server_2, g_server_3, and g_server_4 entries, and then connect to the servers server_1a, server_1b, server_1c, server_1d,server_2, server_3, and server_4 that are in those groups.
    • CONNECT TO @report_1 and CONNECT TO @report_2 connection requests are directed to the replication server that has the lowest transaction latency.
    • The connection between connection_manager_1 and the primary server is prioritized over the connection between connection_manager_2 and the primary server. Failover that would break the connectivity between connection_manager_1 and the primary server is blocked.
    • If failover processing fails after eight attempts, $ONEDB_HOME/etc/CMALARMPROGRAM.sh is called.
    Certain parameters and attributes are not included in this configuration file, so the Connection Manager has the following default behavior:
    • The EVENT_TIMEOUT parameter is not set, so the Connection Managers wait 60 seconds for primary-server events before failover processing begins. The SECONDARY_EVENT_TIMEOUT parameter is not set, so the Connection Managers wait 60 seconds for secondary-server events before the Connection Manager disconnects from the secondary server.
    • The HOST, NETTYPE, SERVICE, and SQLHOSTSOPT attributes of the SLA parameters are not set, so each Connection Manager uses connection information in local and remote sqlhosts files.
    • The SQLHOSTS parameter is not set, so each Connection Manager first searches its local sqlhosts file, and then remote database server sqlhosts files for connectivity information related to server_1, server_2, server_3, and server_4.
    • The WORKERS attributes of the SLA parameters are not set, so four worker threads are allocated to each of the SLAs.
  5. Set the onconfig file DRAUTO configuration parameter on server_1a, server_1b, server_1c, and server_1d to 3
    DRAUTO 3
    This setting specifies that a Connection Manager controls failover arbitration.
  6. Set the onconfig file HA_FOC_ORDER configuration parameter on server_1a to SDS,HDR,RSS
    HA_FOC_ORDER SDS,HDR,RSS

    After the Connection Managers start, and connect to server_1a, the HA_FOC_ORDER value replaces the value of the ORDER attributes in each Connection Manager's configuration file.

    If server_1a fails, the Connection Managers attempt failover to the SD secondary server. If the SD secondary server is also unavailable, the Connection Managers attempt failover to the HDR secondary server. If the HDR secondary server is also unavailable, the Connection Managers attempt failover to the RS secondary server.

  7. Add entries to thesqlhosts files on the hosts of each database server and Connection Manager.
    #dbservername  nettype   hostname  servicename  options
     g_server_1    group     -         -            i=1,c=1,e=server_1d
     server_1a     onsoctcp  host_1    port_1       g=g_server_1
     server_1b     onsoctcp  host_1    port_2       g=g_server_1
     server_1c     onsoctcp  host_2    port_3       g=g_server_1
     server_1d     onsoctcp  host_3    port_4       g=g_server_1
    
     g_server_2    group     -         -            i=2,e=server_2
     server_2      onsoctcp  host_4    port_5       g=g_server_2
    
    g_server_3    group     -          -            i=3,e=server_3
     server_3      onsoctcp  host_5    port_6       g=g_server_3
    
     g_server_4    group     -         -            i=4,e=server_4
     server_4      onsoctcp  host_6    port_7       g=g_server_4 
  8. Create a sqlhosts file on each client host.
    #dbservername  nettype    hostname   servicename   options
     report        group      -          -             c=1,e=report_2
     report_1      onsoctcp   cm_host_1  cm_port_3     g=report
     report_2      onsoctcp   cm_host_2  cm_port_4     g=report
    

    If a Connection Manager fails, client applications can still connect to the other Connection Manager because the report group is defined.

    CONNECT TO @report connection requests are directed through one of the Connection Managers to the replication server that has the lowest transaction latency.

  9. Set each ONEDB_ SQLHOSTS environment variable to the sqlhosts file location by running the setenv command on each Connection Manager and client host.
    setenv ONEDB_ SQLHOSTS path_and_file_name
  10. Turn on quality of data (QOD) monitoring by running the cdr define qod command.
    cdr define qod -c server_1a --start
    The command connects to server_1a, defines server_1a as a master server for monitoring data, and then turns on quality of data monitoring.

    server_1a monitors transaction latency for the replication servers in the grid.

  11. Run the oncmsm utility on each Connection Manager host, to start each Connection Manager.
    On the host of connection_manager_1:
    oncmsm -c cm_1.cfg
    On the host of connection_manager_2:
    oncmsm -c cm_2.cfg
  12. Check each Connection Manager's log file to verify that the Connection Manager started correctly.