OpenTP1 Version 7 Operation
The message queue monitoring facility monitors the backlog of messages in the input queue at regular intervals. This facility can only be applied to user servers (MHPs). The following figure shows an overview of the message queue monitoring facility.
Figure 5-1 Overview of message queue monitoring facility
- After the MCF service starts, the message queue monitoring facility begins monitoring the messages remaining in the input queue.
The phase during which the facility monitors only the number of messages in the message queue is called the queue-depth monitoring phase. Monitoring continues until the MCF service terminates.
- Once the number of messages held in the input queue exceeds a threshold, the facility regularly gauges the ability of the MHP to process those messages.
This is called the processing-capacity judgment phase. During this phase, if the message queue monitoring facility determines that the MHP is unable to process the messages at the expected rate, it either outputs the message KFCA11820-W and continues processing, or outputs the message KFCA11821-E and shuts down the OpenTP1 system.
If a service group defined as a monitoring target is not defined in the attribute definitions for the MCF application, the message queue monitoring facility outputs the message KFCA11822-W and continues processing.
- Organization of this section
- (1) Required operands
- (2) Flow of message queue monitoring
- (3) Example of message queue monitoring
- (4) Notes
(1) Required operands
To use the message queue monitoring facility, specify the following iterations of the mcfmsvg definition command in the MCF manager definition.
- mcfmsvg -g "servgrpn=service-group-name"
Specify the name of the service group whose input queue you want to monitor.
- mcfmsvg -w "watchcnt=input-queue-message-count-threshold"
Specify the threshold for the number of messages that can reside in the queue before the message queue monitoring facility transitions from monitoring the queue depth to judging processing capacity.
- mcfmsvg -w "watchint=input-queue-monitoring-interval"
Specify the interval at which the message queue monitoring facility checks message numbers and judges the MHP processing capacity.
- mcfmsvg -w "expectcnt=expected-number-of-service-requests-processed-by-MHP"
Specify the number of service requests that the MHP is expected to have processed by the next time that processing capacity is judged (the number of requests that must be processed to clear the messages in the input queue).
- mcfmsvg -w "abort=yes|no"
Specify whether to shut down the OpenTP1 system when MHP is judged to have insufficient processing capacity.
For details about the operands, see the manual OpenTP1 System Definition.
(2) Flow of message queue monitoring
The processing to monitor the backlog of queued messages is as follows:
- When the MCF service is started, the message queue monitoring facility starts in the queue-depth monitoring phase, and begins monitoring the number of messages in the input queue at the specified interval (as specified by mcfmsvg -w watchint).
- When the number of messages in the input queue exceeds the number specified by mcfmsvg -w watchcnt, the message queue monitoring facility enters the processing-capacity judgment phase.
While in the processing-capacity judgment phase, the message queue monitoring facility uses the following formula to determine the processing capacity of the MHP.
- Formula for determining MHP processing capacity:
Depending on the outcome, the following processing takes place once this judgment is made.
- When the formula for determining MHP processing capacity does not hold true:
OpenTP1 continues processing.
- When the formula holds true, and the OpenTP1 system is not set to shut down when MHP processing capacity is deemed inadequate (no is specified for mcfmsvg -w abort):
The message KFCA11820-W is output and OpenTP1 continues processing.
- When the formula holds true, and the OpenTP1 system is set to shut down when MHP processing capacity is deemed inadequate (yes is specified for mcfmsvg -w abort):
The message KFCA11821-E is output and the OpenTP1 system is shut down.
When the number of messages held in the input queue falls below the number specified by mcfmsvg -w watchcnt, the message queue monitoring facility leaves the processing-capacity judgment phase and re-enters the queue-depth monitoring phase.
The following table shows the conditions judged during message queue monitoring, and how MCF behaves in each instance.
Table 5-2 Relationship between judgment conditions during message queue monitoring and MCF behavior
Judgment condition |
MCF behavior |
Previous judgment phase |
Relation between number in queue and threshold |
Relation between A and B |
Relation between A and C |
mcfmsvg -w abort specification |
Queue-depth monitoring phase |
Number in queue < Threshold |
N/A |
N/A |
N/A |
Stays in queue-depth monitoring phase and continues processing. |
Number in queue Threshold |
N/A |
N/A |
N/A |
Enters processing-capacity judgment phase and continues processing. |
Processing-capacity judgment phase |
Number in queue < Threshold |
N/A |
N/A |
N/A |
Enters queue-depth monitoring phase and continues processing. |
Number in queue Threshold |
A B |
A C |
N/A |
Remains in processing-capacity judgment phase and continues processing. |
A < C |
N/A |
A < B |
A C |
N/A |
A < C |
Yes |
Outputs message KFCA11821-E, forcibly terminates MHP, and shuts down OpenTP1 system. |
No |
Outputs message KFCA11820-W, remains in processing-capacity judgment phase and continues processing. |
- Legend:
- A: Number of service requests processed by MHP
- B: Number of service requests expected to be processed by MHP
- C: Number of service requests queued at time of previous judgment
(3) Example of message queue monitoring
The following is an example of the processing of the message queue monitoring facility, assuming a situation in which the mcfmsvg definition command is specified as follows in the MCF manager definition:
- Specification of mcfmsvg definition command in MCF manager definition:
Because 30 is specified as the threshold for messages in the input queue, the message queue monitoring facility will enter the queue-depth monitoring phase for segments C2 to C5, and from C8 onward, in the above figure. For all other segments, the message queue monitoring facility will be in the processing-capacity judgment phase.
The following table shows the relationship between the number of messages in the input queue when the queue status is judged, and the judgment result.
Table 5-3 Queued message counts and judgment results
No. |
Number of queued messages carried over from previous judgment
(Bn) |
Number of queued messages at previous judgment
(Pn-1) |
Number of service requests processed since last judgment
(Pn-1 - Bn) |
Number of service requests expected to be processed by MHP |
Judgment result |
1 |
11 |
- |
- |
24 |
Because the number of queued messages has not reached the threshold, the queue-depth monitoring phase remains in effect and OpenTP1 stays online. |
2 |
9 |
18 |
9 |
24 |
3 |
25 |
28 |
3 |
24 |
Because the number of queued messages has reached the threshold, the processing-capacity judgment phase begins. |
4 |
8 |
32 |
24 |
24 |
Although the number of queued messages has remained above the threshold, the MHP was able to process more than the expected number of service requests. Thus, the processing-capacity judgment phase remains in effect and OpenTP1 stays online. |
5 |
13 |
45 |
32 |
24 |
6 |
0 |
35 |
35 |
24 |
7 |
3 |
30 |
27 |
24 |
Because the number of queued messages has fallen below the threshold, the queue-depth monitoring phase begins. |
8 |
5 |
11 |
6 |
24 |
Because the number of queued messages has not reached the threshold, the queue-depth monitoring phase remains in effect and OpenTP1 stays online. |
9 |
15 |
17 |
2 |
24 |
Because the number of queued messages has reached the threshold, the processing-capacity judgment phase begins. |
10 |
29 |
32 |
3 |
24 |
Because the number of queued messages has remained above the threshold, and the MHP was unable to process the expected number of service requests, OpenTP1 is taken offline. |
- The message queue monitoring facility does not monitor service groups that are shut down or being held. Service groups that enter shutdown or held status during the processing-capacity judgment phase are assumed to have transitioned to the queue-depth monitoring phase. Accordingly, for the first monitoring cycle after the service group is released from shutdown or held status, the message queue monitoring facility monitors queue depth regardless of the phase that was in effect when the group was shut down or held. If the number of messages in the queue is found to exceed the threshold, the message queue monitoring facility will then begin the processing-capacity judgment phase.
- The messages in input queues that are deleted by the mcftdlqsg command are not counted in the number of service requests processed by the MHP.
- Do not advance or set back the time by whole seconds while OpenTP1 is online. Because this facility uses data and time information to operate, changing the time while OpenTP1 is online may cause the following problems:
- If you advance the time, the MCF may wrongly determine that the time to judge the queue status or processing ability has arrived and, depending on the definitions in effect, may shut down the OpenTP1 system.
- If you set the time back, depending on the definitions in effect, the MCF may not monitor queue depth.
For details about changing the time, see 3.13.4 Notes on changing the time.
All Rights Reserved. Copyright (C) 2006, 2010, Hitachi, Ltd.