PRIMECLUSTER Installation and Administration Guide4.6 Cloud Services

21.10 Building Cluster Application

For details on how to build the cluster application, refer to "Chapter 6 Building Cluster Applications" in "PRIMECLUSTER Installation and Administration Guide."

When using the mirroring among servers, set the mirroring among the servers of GDS (creating netmirror volume) while building the cluster application.

Also, depending on the architectural pattern selected for network takeover in "20.2.1 Network Takeover", network takeover needs to be registered in the cluster application. To build a cluster application for network takeover, refer to "21.10.1 Building Cluster Application for Network Takeover."

Note

• Note the following points when performing the settings in "Setting Tuning Parameters" in "PRIMECLUSTER Global Disk Services Configuration and Administration Guide."

  • Among the tuning parameters to be set, set the following values for the following tuning parameters.

    Tuning parameter name	Value after change
    ED_CMD_RETRY_COUNT	100
    ED_DRV_RETRY_COUNT	100

    Example:

    ED_CMD_RETRY_COUNT=100
    ED_DRV_RETRY_COUNT=100

    To extend the timeout period (CLUSTER_TIMEOUT) of the CF heartbeat, change the above parameter values according to the following formula. Round up the values after the decimal point.

    Calculation formula:

    <Increased CLUSTER_TIMEOUT> / 3 + 100

  • When using the asynchronous forcible stop method, set the following tuning parameter in /etc/opt/FJSVsdx/sdx.cf.

    SDX_NETMIRROR_IO_BLOCKADE=1

  • When using the asynchronous forcible stop method, comment out the following tuning parameter in the /etc/opt/FJSVsdx/modules/sfdsk.conf and disable the slice preceding degenerated option.

    SDX_NETMIRROR_PRE_DETACH=1;

    Example:

    #SDX_NETMIRROR_PRE_DETACH=1;

• Note the following setting in "Checking and Setting Required Packages" in "PRIMECLUSTER Global Disk Services Configuration and Administration Guide."

  • When using the asynchronous forcible stop method, add "smawcf.service" to the value "After" in /etc/systemd/system/fjsvsdx.service.d/netmirror.conf.

    Example:

    After=target.service smawcf.service

• If the icmp communication between cluster nodes is not allowed in the security group configuration, the following message is displayed when the clchkcluster command is executed.

  Admin IP <IP address> used by SF is not alive.

  If this message is output, refer to "20.3.2.1 Rules Applied to the Administrative LAN", and set the icmp protocol rule to allow the icmp communication between cluster nodes. After that, execute the clchkcluster command again.

• When using the Symfoware Server (Native) Hot Standby feature, the standby and scalable userApplication configurations are required.

  Set the standby userApplication attribute to the following value.

  Attribute	Setup value
  StandbyTransitions	ClearFaultRequest | SwitchRequest

  Refer to "6) Scalable operation" in "6.7 Setting Up Cluster Applications" in the "PRIMECLUSTER Installation and Administration Guide" for detailed build instructions.