Hitachi

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


4.1.1 HA Monitor's control of server startup

The user starts a server in the server mode by executing the start command provided by the program, and starts a server in the monitor mode by executing the monitor-mode server start command (monbegin command).

When a server is started, HA Monitor checks the server environment definition and the server status on the remote host to determine whether the server is to be started as the active server or a standby server. This subsection explains how HA Monitor determines the active server or standby server. It also explains the HA Monitor processing that occurs when the server status on the remote host cannot be checked due to a monitoring path failure.

Organization of this subsection

(1) Determining the active server or standby server during normal operation

Normally, HA Monitor determines whether a server is to be started as the active server or a standby server on the basis of the start type specified in the initial operand in the server environment definition and the status of the remote host. The following table describes how this determination is made.

Table 4‒1: Determining whether the active server or a standby server

Status of remote host

Server environment definition

Defined as an active server

Defined as a standby server#1

Active server exists.

Startup processing is underway.

Start as a standby server.

Start as a standby server.

Execution processing is underway.

Start as a standby server.

Start as a standby server.

Termination processing is underway.

Cannot be started.#2

Cannot be started.#2

Standby server exists.

Startup processing is underway.

Start as the active server.

Cannot be started.#2

Execution processing is underway.

Start as the active server.

Cannot be started.#2

Termination processing is underway.

Start as the active server.

Wait until the active server starts.#3

Waiting for grouped-system switchover.#4

Start as the active server.

Start as the active server.

Hot standby switching from standby server to active server is underway.

Start as a standby server.

Start as a standby server.

Retrying startup.#5

Start as the active server.

Start as a standby server.

No server is found (HA Monitor is active).

Start as the active server.

Wait until the active server starts.#3

No server is found (HA Monitor is inactive).

Start as the active server.#7

Wait until the active server starts.#3

The server status cannot be checked.

Retry processing.#6

Wait until the active server starts.#3

#1: If the status is Retrying startup, the operation is the same as for Defined as an active server.

#2: Cancels the server startup processing.

#3: Places startup of the standby server on hold until startup of the active server on the remote host is confirmed (active server start wait state). After start of the active server startup processing is confirmed, HA Monitor starts the server as the standby server.

#4: For details about the grouped-system switchover wait state, see 4.4.1 Server switchover type during grouped-system switchover.

#5: For details about the retrying startup status, see 3.2.7 Performing start retries during hot standby processing.

#6: If HA Monitor is unable to check the server status on the remote host due to a host failure or a monitoring path failure on the remote host, it retries for only one minute. If the status still cannot be checked after one minute of retries, HA Monitor determines that no server is available on the remote host due to a failure and places the server in the active server start wait state.

#7: If it is impossible to check the status of a host for which pair is specified in the connection configuration file, HA Monitor places the server in the active server start wait state.

The method used to determine the active server or standby server is different when the server defined as the active server is restarted following a failure. For details, see (2) Start type when the server is restarted.

For details about starting and terminating a server in the wait state as the active server, see 7.4.1 Starting a server in the wait state and then restarting jobs.

(2) Start type when the server is restarted

When a server is restarted after a failure, an active server might already be running on the remote host. Therefore, HA Monitor does not start the server as an active server, even if it is defined as an active server, until the server status on the remote host can be verified. Once the server status on the remote host has been verified, HA Monitor determines whether the server is to be started as the active server or a standby server.

The following table describes how the start type is determined when a server defined as the active server is restarted.

Table 4‒2: Determining the start type when the server is restarted

Status of remote host

Server start type#1

Active server exists.

Startup processing is underway.

Start as a standby server.

Execution processing is underway.

Start as a standby server.

Termination processing is underway.

Cannot be started.#2

Standby server exists.

Startup processing is underway.

Start as the active server.

Execution processing is underway.

Cannot be started.#2

Termination processing is underway.

Wait until the active server starts.#3

Waiting for grouped-system switchover.

Start as the active server.

Hot standby switching from standby server to active server is underway.

Start as a standby server.

No server is found (or HA Monitor is inactive).

Wait until the active server starts.#3

The server status cannot be checked.

Wait until the active server starts.#3

#1: The start type specified in the initial operand in the server environment definition is ignored.

#2: Cancels the server startup processing.

#3: Places startup of the standby server on hold until startup of the active server on the remote host is confirmed (active server start wait state). After start of the active server startup processing is confirmed, HA Monitor starts the server as a standby server.

For details about starting and terminating a server in the wait state as the active server, see 7.4.1 Starting a server in the wait state and then restarting jobs.