Subsystem Control Point (SCP) Management Programming Manual

Communicating With an SCP Process

Subsystem Control Point (SCP) Management Programming Manual—520619-001

2-13

Stopping an SCP Process

An SCP process can stop itself. How and when it does this depends on the value of the

AUTOSTOP SCP process execution parameter you specify when you start the SCP

process. See Starting an SCP Process Using the RUN Command

on page 2-2.

An SCP process can also be stopped by a STOP command from a command interpreter

such as TACL, a STOP or PROCESS_STOP_ procedure call, or a STOP or ABORT

SPI command from a management process or SCF process.

For SCP support for the STOP and ABORT commands, see Section 5, Commands and

Responses.

Abnormal Process Termination

If a non-fault-tolerant SCP fails while it is processing a command, the WRITEREAD

call issued by a requesting application returns file-system error 201.

If an application opens a fault-tolerant SCP with sync-depth equal to 0, and if the

primary process is stopped or its CPU fails, the application’s WRITEREAD procedure

call returns file-system error 201.

If an application opens a fault-tolerant SCP process with sync-depth equal to 1, and if

the primary process is stopped or its CPU fails, the backup process takes over, and the

command message is reprocessed. Reprocessing the command can result in errors

because part or all of the command might have executed before the backup takeover.

If a management application starts an SCP process that subsequently stops abnormally,

the Guardian operating system sends system message –101 to the application. For more

information, see the descriptions of the OPEN and FILE_OPEN_ procedures in the

Guardian Procedure Calls Reference Manual and the discussion of system messages in

the Guardian Programmer’s Guide.

In the case of any of these failures, part or all of a pending command might have been

executed. After starting another SCP process, you or your application should check the

status of objects that the command could have affected. Associated error and event

messages may help you determine the cause of the problem.