Hitachi

For Linux(R) (x86) Systems HA Monitor Cluster Software


4.6.2 Processing flow for connecting and disconnecting shared resources when a resource server is used

When a resource server is used, shared resources are connected when the resource server is started and disconnected when the resource server is terminated. HA Monitor controls the order in which the servers and resource server defined as a group are started so that the servers can share the resources.

This subsection explains the order in which the servers and resource server are started, the processing flow for connecting and disconnecting shared resources, and the order in which the servers and resource server are terminated.

How the active server or standby server is determined when a resource server is started is the same as when other servers are started. For details about how to determine an active server or a standby server, see 4.1.1 HA Monitor's control of server startup.

Organization of this subsection

(1) Order in which the servers and resource server are started

HA Monitor treats the servers and resource server in a group as having parent-child relationships and starts them sequentially beginning with the parent. The following figure shows the parent-child relationships and the start order using as an example a server group consisting of server 1, server 2, and a resource server.

Figure 4‒34: Parent-child relationships and start order of a server group that includes a resource server

[Figure]

To use this resource server, the user must specify it as the parent server of servers 1 and 2. A resource server must always be specified at the top of the hierarchy as the parent server. HA Monitor controls the start order based on the specified parent-child relationships in such a manner that the servers are started sequentially starting with the parent server.

(2) Processing flow for connecting and disconnecting shared resources

This subsection explains the processing flow for connecting and disconnecting shared resources when a resource server is used. The explanation covers the period from the time the servers are started to the time the hot standby operation is performed in the event of a failure. For details about the parent-child relationships of the server group explained here, see (1) Order in which the servers and resource server are started. For details about HA Monitor and server processing, see 4.8 Processing flows.

The following figure shows the processing flow for connecting and disconnecting shared resources when a resource server is used.

Figure 4‒35: Processing flow for connecting and disconnecting shared resources (when using a resource server)

[Figure]

The following describes the steps of the connection flow in the preceding figure. The numbers correspond to the numbers in the figure.

  1. When the operator starts server 1 or 2, HA Monitor starts the resource server. At the same time, HA Monitor connects the shared resources. Until the resource server's startup processing is completed, HA Monitor keeps servers 1 and 2 in the active system in the resource server start wait state so that they cannot be started.

  2. Once the shared resources have been connected in the active system, HA Monitor completes startup processing for servers 1 and 2.

    If the operator starts server 1 or 2 in the standby system, HA Monitor starts the resource server.

  3. If a server failure occurs, HA Monitor performs a grouped-system switchover. In this example, HA Monitor terminates servers 1 and 2 before terminating the resource server because a server failure has occurred on server 1.

  4. When servers 1 and 2 in the active system have terminated, HA Monitor terminates the resource server. At the same time, HA Monitor disconnects the shared resources.

    When the shared resources have been disconnected in the active system, the resource server in the standby system starts the hot standby operation. When the hot standby operation starts, the shared resources are connected. In this case, servers 1 and 2 in the standby system are placed in the hot-standby completion wait state.

    If you use the server and host status display command (monshow command), (ONL) is displayed as the status of servers 1 and 2. (ONL) means that the active servers are engaged in startup processing, but the servers themselves are actually waiting for completion of the hot standby operation on the resource server. In other words, the status displayed by the command is not the same as the actual server status.

  5. After connecting the shared resources, HA Monitor starts servers 1 and 2 in the standby system as active servers. If the operator restarts server 1 or 2 in the active system, the resource server and servers 1 and 2 are started as standby servers.

(3) Order in which the servers and resource server are terminated

A resource server is terminated automatically when all the servers for whom the resource server is the parent server have terminated. The resource server is not terminated automatically when a server in the same group has been terminated if its parent server is not the resource server. If a resource server is terminated in the active system, the resource server in the corresponding standby system is also terminated automatically.