I have an app that does very long tasks like this a few times a week.
The tasks I have a *very* EOF intensive, as in hammering the database
usually for a very long time (4 to 24 hours)
Here is approach I use:
Create a Callable (or Runnable) class for the task.
Create setters/getters for any initial variables you need to pass in.
For EnterpriseObjects, just pass in the EOGlobalIDs, not the objects
themselves.
In the task method itself:
- Create a new ObjectStoreCoordinator (use Wonder OSC synchronization
if you want changes to propogate)
- Create any editing contexts using that single task OSC as a parent
- lock and unlock your ec's manually when working with them
- depending on how much EO's you are fiddling with, memory management
can be challenging, so recycling ec's, and, dare I say it, resorting
to checking available memory and forcing garbage collection when
memory almost exhausted (some will cringe at this, but extreme EOF
just sucks memory and does not let go fast enough ... depending on the
conditions)
Tips
- Surround the task with a try/catch to get any errors
- Send an email to the user (if user initiated), or admin (if
necessary) notifying them of success, failure or error when task is
done (or error is thrown).
- Using log4j smtp appender for ERROR level is good too to ensure
errors in anonymous tasks are alerted to admin/devs promptly.
On Jul 30, 2008, at 12:23 AM, Jeff Schmitz wrote:
Hello,
I have the need to kick off a VERY long (i.e. possibly up to 8-10
hour) background task that before it's through fetches most of the
contents of the database (via EO's of course) and performs
calculations on the data and saves the calculated values back to the
EOs many times during the process. And oh, I'd like to have the
results available to the rest of my app as they are saved by the
background task.
Currently btw, I do this with a java thread and don't really use
EOs, and I just return a page immediately after the thread is kicked
off, i.e. I don't really care about a status page reloading for 8
hours as I'm the only user that actually kicks off this process.
My first question is, now that I'm using EOs throughout my app
(including the background process) is there any reason to use the
WOLongResponsePage if I still don't care about the status page? If
so, there seems to be a dearth of information on exactly how to use
such a component. All I can really find is a very terse API doc,
and a few mentions of its existance on the wiki. Any examples out
there anywhere?
I'll save my context locking and memory flushing questions until
after I've researched the subject a little more. From what I can
tell though, I'll want to create a new editing context for the
background thread (not sure about needing an independent Object
store coordinator for the background process, but I don't think so),
so that's a start.
Finally, anything in project Wonder that can help me? I did run
across this:
http://webobjects.mdimension.com/wonder/api/er/extensions/concurrency/class-use/ERXLongResponseTask.html#er.extensions.concurrency
but again, very terse docs.
Thanks,
Jeff
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/webobjects-dev/kieran_lists%40mac.com
This email sent to [EMAIL PROTECTED]
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com
This email sent to [EMAIL PROTECTED]
