3.7.4 Locking and controlling requests for a schedule queue

(1) Overview of locking and controlling requests for a schedule queue

You can directly lock a schedule queue. This allows you to simultaneously stop all J2EE applications that share the schedule queue. You can select whether to discard requests remaining in the schedule queue immediately, or continue processing them for a certain length of time. If you choose to continue processing, you can specify the timeout value to forcibly discard the requests that are not processed in the specified time. Processing of requests that are being executed on J2EE servers continues.

In response to a request to lock a schedule queue, the CTM daemon closes the exit of the schedule queue to perform a service lock so that the queue receives no more requests. Already-queued requests are handled (discarded or processed) as pre-specified, and then the lock of the schedule queue is completed. If queued requests are discarded, they are returned to the clients as errors. If queued requests are processed, after processing continues for a preset time, any requests whose processing does not end within the time are returned as errors.

The following figure shows an overview of locking and controlling requests for a schedule queue.

Figure 3‒18: Overview of locking and controlling requests for a schedule queue

In a back-end system that uses CTM, to simultaneously stop all J2EE applications on a host or all J2EE applications that share a schedule queue, directly lock the schedule queue for those J2EE applications. You can then stop the J2EE applications.

If you use the management command to directly lock schedule queues for J2EE applications, you can stop J2EE applications for each schedule queue that is shared by J2EE applications, for each host, or for each management domain. When you do so, you can select whether queued requests are to be discarded immediately or processed for a certain length of time. Any queued requests that you choose to discard are returned to the clients as errors. If you choose to process queued requests for a certain length of time, requests whose processing does not end in the specified time are returned to the clients as errors.

How you can use the management command to perform a CTM service lock is described below. The following subsections describe two types of locks: forced locking and timeout-triggered locking. Use forced locking in cases such as when you must lock a queue immediately because the CTM daemon is in a high load state.

To Page Top

(2) Forced locking of a schedule queue

You can lock a schedule queue without it communicating with the CTM daemon. This is called forced locking of a schedule queue. Forced locking can be used to immediately lock a queue in such a case where the load on the CTM daemon is high. If you use normal locking, the schedule queue communicates with the CTM daemon, and then discards the requests in the queue. In this case, if the CTM daemon is in a high-load state, communication with the CTM daemon takes time. Therefore, time is also needed before the queue is locked.

If you use forced locking, you can immediately lock the queue because communication between the queue and the CTM daemon is skipped. Note that discarding of queued requests takes place the next time the CTM daemon exchanges the load information with other CTM daemons.

To perform forced locking, execute the mngsvrutil management command with the hold subcommand and queue force argument specified. When you perform forced locking of a schedule queue, the requests in the queue are discarded after a certain length of time. If you do not want to discard the requests, execute the ctmholdque command with the -CTMRequestLeave option specified.

You can unlock a schedule queue locked by forced locking in the same way as you unlock a schedule queue locked in the ordinary manner. For details about the mngsvrutil command, see mngsvrutil (Management Server management command) in the uCosminexus Application Server Command Reference Guide.

The following shows the format and an execution example of the mngsvrutil command executed to perform forced locking.

Format:

mngsvrutil -m Management-Server-host-name[:port-number] -u management-user-ID -p management-password -t host-name -k host hold queues force

Execution example:

mngsvrutil -m mnghost -u user01 -p pw1 -t host01 -k host hold queues force

To Page Top

(3) Timeout-triggered locking of a schedule queue

The CTM daemon of a schedule queue monitors EJB client timeouts at regular intervals, and locks the schedule queue when the number of timeout occurrences exceeds a preset value. This is called timeout-triggered locking of a schedule queue.

The following describes timeout occurrences. First, see the following figure.

Figure 3‒19: How timeout-triggered locking of a schedule queue occurs

In the above figure, the number of timeout occurrences is watched every 10 seconds. Counting of timeout occurrences continues only within each watcher interval. When a watcher interval ends, the counter is reset, and the next watcher interval starts.

For example, assume that the timeout count threshold is set to 10. In this case, if 10 or more timeouts occur in a 10-second watcher interval, the queue is locked. Note that if 10 timeout occurrences are detected, locking of a queue takes place when the next watcher interval starts. In the above figure, 11 timeouts are detected 30 seconds after the start of watching. Therefore, the queue is locked after 30 seconds has elapsed since watching started.

Timeout-triggered locking for a schedule queue is set by using an option specified during startup of the CTM daemon. Specify the -CTMWatchRequest option when executing the ctmstart command.

To Page Top