GitHub - grtjn/ml-queue: Background processing queue for MarkLogic Server

grtjn / ml-queue Public

Notifications You must be signed in to change notification settings
Fork 0
Star 5

Background processing queue for MarkLogic Server

Notifications

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
README		README
default.xqy		default.xqy
do-queue-action.xqy		do-queue-action.xqy
queue-cron.xqy		queue-cron.xqy
queue-lib.xqy		queue-lib.xqy
task-a.xqy		task-a.xqy
task-b.xqy		task-b.xqy
task-c.xqy		task-c.xqy
view-task.xqy		view-task.xqy

Repository files navigation

Background processing queue for MarkLogic Server

Description
-----------

The normal task server queue doesn't discriminate between kinds of tasks. CPF processing can for instance flood the queue, and cause front-end request respond slowly. This queue is maintained separately, allows prioritizing, and the cron is designed only to launch tasks when the task server has spare threads. In other words, ideal for background processing.

This code was written as an example, has not been tested in real life. Feedback is welcome.

Installation
------------

- Drop the contents of this folder anywhere, where it is accessible by MarkLogic Server. So upload it to Modules, or drop in below a folder that has been specified as root of one of your app servers. Personally, I tested it in a subfolder called /tasks/ of the root folder of one my HTTP app servers.
- Access the code through http://localhost:nnnn/path-to-ml-queue/ (for instance in my case http://localhost:8000/tasks/)
- It shows the Task queue manager. The top shows some usefull utilities, and details. Then the option to create a task. At the bottom the current queue.
- As example create the following tasks: /path-to-ml-queue/task-a.xqy, /path-to-ml-queue/task-b.xqy, /path-to-ml-queue/task-c.xqy (default prio)
- The queue initially shows them in creation order.
- Hit + behind task a, it will move to the bottom of the queue.
- Hit - behind task c, it will move to the top of the queue.
- Now you can hit 'start cron'. After a 10 sec delay, the manager is shown again, stating the cron is active. First task (being task c) should have already been taken from the queue.
- Hit 'auto refresh' near the top to follow progress with a 5 sec refresh rate. Cron sleeps at intervals of 1 minute.
- Hit 'stop cron' once the queue is empty. (Hit 'auto refresh' again to follow progress, cron should deactivate in max 1 minute.)

The cron process and the three tasks write log messages. Check ErrorLog.txt, it should report messages like follows:

2011-10-15 16:14:21.540 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:14:26.549 Info: TaskServer: Task C!
2011-10-15 16:14:26.559 Info: TaskServer: Queue cron 14311129094746556270: launched task 56902235478346153 2011-10-15T16:12:17.889+02:00 /tasks/task-c.xqy -1..
2011-10-15 16:16:36.643 Info: TaskServer: Queue cron 14311129094746556270: sleeping 60000 msec..
2011-10-15 16:15:26.566 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:15:31.629 Info: TaskServer: Queue cron 14311129094746556270: launched task 16267903407942429866 2011-10-15T16:12:13.286+02:00 /tasks/task-b.xqy 0..
2011-10-15 16:15:31.629 Info: TaskServer: Task B!
2011-10-15 16:16:36.643 Info: TaskServer: Queue cron 14311129094746556270: sleeping 60000 msec..
2011-10-15 16:16:31.634 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:16:36.643 Info: TaskServer: Queue cron 14311129094746556270: launched task 14720630703100337358 2011-10-15T16:12:07.66+02:00 /tasks/task-a.xqy 1..
2011-10-15 16:16:36.643 Info: TaskServer: Queue cron 14311129094746556270: sleeping 60000 msec..
2011-10-15 16:16:36.643 Info: TaskServer: Task A!
2011-10-15 16:17:36.645 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:17:41.654 Info: TaskServer: Queue cron 14311129094746556270: nothing to do..
2011-10-15 16:17:41.654 Info: TaskServer: Queue cron 14311129094746556270: sleeping 60000 msec..
2011-10-15 16:18:41.657 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:18:46.665 Info: TaskServer: Queue cron 14311129094746556270: nothing to do..
2011-10-15 16:18:46.665 Info: TaskServer: Queue cron 14311129094746556270: sleeping 60000 msec..
2011-10-15 16:19:46.672 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:19:51.681 Info: TaskServer: Queue cron 14311129094746556270: nothing to do..
2011-10-15 16:19:51.681 Info: TaskServer: Queue cron 14311129094746556270: sleeping 60000 msec..
2011-10-15 16:09:21.793 Info: TaskServer: Queue cron 14311129094746556270: beginning..
2011-10-15 16:09:26.794 Info: TaskServer: Queue cron 14311129094746556270: stopping..

NOTE: I changed the function q:get-task-server-threads-available to always return 1 to make the above example more descriptive. Uncomment the original code that have multiple tasks launched at each iteration.

Limitations
-----------
The tasks do not take parameters yet, nor can you specify spawn options. Patches welcome!

More importantly, the spawn recursion depth is limited (thnx to MarkLogic experts for pointing that out to me). So I tested and the recursion depth is exactly 1000. This happens to coincide with the 'pre-commit trigger depth' that can be configured on each App server, or perhaps that isn't a coincidence at all (haven't tested yet). To circumvent this, use a Scheduled task (found under Group configuration in the MarkLogic Server Admin interface).