OpenTP1 Version 7 Description
OAs with SPPs, OpenTP1 schedules messages to MHPs by creating a schedule queue for each MHP service group.
First MCF converts the MCF-application name included in the message into the service group name and the service names, and then catalogs the received message into the input queue. When the last segment of a received message is stored in the input queue, MCF catalogs the messages into the proper schedule queue depending on the correspondence between the MCF application and the service group name and service names defined in the application attribute definition. Scheduling fails unless the MHP service group and the MCF that received the message are located on the same node.
Messages to an MHP are taken out from the schedule queue by the FIFO method.
Figure 3-46 shows MHP scheduling.
Figure 3-46 Scheduling messages for an MHP
An OpenTP1 administrator can use commands to shut down scheduling of service requests to an MHP, and to release such scheduling shutdowns. As shown in the following table, these commands can apply to the scheduling of service requests to MCF applications, services, or service groups.
The user can specify whether the service group will become active or remain shut down after complete recovery from an abnormal termination. The user can make this decision either when the user uses the mcftdctsg command to shut down the service group, or when using the status inheritance definition to create the system.
Table 3-9 Shutting down scheduling of service requests to MCF-applications
Type of scheduling shutdown | Processing of messages that arrived after the shutdown | Processing of messages already cataloged in the input queue before the shutdown |
---|---|---|
Preventing both entry to input queue and output from schedule queue | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is recataloged in the input queue for the MHP for ERREVT2 (an MCF event). |
Preventing entry to input queue only | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The relevant MHP is started normally. |
Preventing only scheduling | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is recataloged in the input queue for the MHP for ERREVT2 (an MCF event). |
Figure 3-47 Shutting down scheduling by specifying an MCF application name
Table 3-10 Shutting down scheduling of service requests to MHP services
Type of scheduling shutdown | Processing for messages that arrived after the shutdown | Processing for messages already cataloged in the input queue before the shutdown |
---|---|---|
Preventing both entry to input queue and output from schedule queue | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is recataloged in the input queue for the MHP for ERREVT2 (an MCF event). |
Preventing entry to input queue only | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The MHP is started normally. |
Preventing only scheduling | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is recataloged in the input queue for the MHP for ERREVT2 (an MCF event). |
Table 3-11 Shutdown of MHP service group
Type of scheduling shutdown | Processing for messages that arrived after the shutdown | Processing for messages already cataloged in the input queue before the shutdown | |
---|---|---|---|
Input queue is disk queue | Preventing both entry to input queue and output from schedule queue | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is placed in the schedule wait state. After the scheduling shutdown is released, the MHP is started normally. |
Preventing entry to input queue only | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The MHP is started normally. | |
Preventing output from schedule queue only | The message is cataloged in the input queue and the message is placed in the schedule wait state.# | The message is placed in the schedule wait state. After the scheduling shutdown is released, the MHP is started normally. | |
Input queue is memory queue | Preventing both entry to input queue and output from schedule queue | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). |
Preventing entry to input queue only | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The MHP is started normally. | |
Preventing output from schedule queue only | The message is cataloged in the input queue for the MHP for ERREVT2 (an MCF event). | The message is recataloged in the input queue for the MHP for ERREVT2 (an MCF event). |
In the application attribute definition, you can define whether or not scheduling for an MCF application or service should be shut down if an MHP terminates abnormally while executing a service. Also, in the application attribute definition, you can define that scheduling for an MCF application or service should be shut down after a specified number of abnormal terminations of the MHP. You can specify whether the number of abnormal terminations refers to the number of consecutive MHP abnormal terminations, or the number of MHP abnormal terminations in total.
To shut down in units of applications, use the aplihold command in the application attribute definition. To shut down in units of services, use the servhold command in the application attribute definition. To shut down in units of service groups, use the srvghold command in the application attribute definition.
Note that an abnormal termination includes the situation in which an MHP is rolled back by the rollback function dc_mcf_rollback() that has no-return specified in a parameter. For details of the rollback function, see the OpenTP1 Programming Guide.
When shutting down only the scheduling in units of service groups, you can choose how to handle the received messages if MHP terminates abnormally while executing a service. You can:
For automatic shutdown due to abnormal termination, the shutdown status of MHP is not inherited at a full recovery operation after a stoppage of the online system.
When the specification that prevents shutdown at abnormal termination is specified by aplihold, servhold, or srvghold in the application attribute definition, the application, service, or service group will not be shut down. If frequent abnormal terminations occur with this specification, use a command to shut down the application, service, or service group.
When MCF schedules an MHP, it checks for a service group that corresponds to the application name specified in the application attribute definitions. If no such service group exists, OpenTP1 shuts down scheduling for that service group only, and outputs a message to the message log to report that there is no corresponding service group. See Table 3-11 for differences between situations where the input queue is a disk queue and where the input queue is a memory queue.
The MHP scheduling-shutdown status, when scheduling for the MHP was automatically shut down because of an inconsistency between the MCF-application name and the MHP, is not inherited during a full recovery operation after a stoppage of the online system.
All Rights Reserved. Copyright (C) 2006, 2010, Hitachi, Ltd.