Re: Warning: Request will be REDIRECTED to non-refusing instance
Re: Warning: Request will be REDIRECTED to non-refusing instance
- Subject: Re: Warning: Request will be REDIRECTED to non-refusing instance
- From: Chuck Hill <email@hidden>
- Date: Wed, 24 Sep 2003 09:33:39 -0700
Hi Daniel,
At 10:28 AM 24/09/2003 +0200, Daniel Muller wrote:
>On Mercredi, sep 24, 2003, at 00:59, Chuck Hill wrote:
>
>> It sounds like you may have a deadlocking problem in your app
>> (instance 1 in the explanation below).
>
>Hi Chuck,
>
>After more investigation, the WO Adaptor receives the request and send
>to application's instance and wait 45 sec. for a reply.
>The instance never respond but wotaskd is able to shutdown it.
>
Is that the first request? If so, you can try increasing the app launch
time in JavaMonitor (I forget what it is called exactly). Also, if you are
restarting all 10 apps at once that might be too much. Try doing 5 then
another 5 an hour later or something like that. This will also prevent
service outages.
>I've changed
>the minimum sessions to 10 like that wotaskd is able to restart
>instances every day.
>
That will kill ten active users...
>> If you can get a thread dump (see the archives here or on Omnigroup)
>> that should show if this is the case.
>
>Seems to be used only from Terminal, in my case the application is
>always working fine during development
>it's only while running on deployment that this problem happened.
>
>Is it possible to get a thread dump of deployed instance?
>
Deployed on what? You can do it with a kill signal on UNIX. On Windows,
there is no way. Another, perhaps better, alternative is it launch the
apps so that you can connect with a remote debugger. Then you can get all
the info you need.
>> Are you being careful to lock and unlock all of your editing contexts
>> correctly?
>
>Yes I do, but I never lock/unlock the defaultEditingContext.
>I'm wondering if it's required in some context?
That is locked for you as long as you only access it between the calls to
awake() and sleep() in session. You ***do*** need to lock any editing
contexts *nested* in this one (child ECs).
>BTW, I'm using Oracle database and I can see in the ora trace log that
>there are deadlocks.
That could also be the problem. Or it might just be a symptom. Perhaps
the isolation level is set incorrectly?
>EOEditingContext lock/unlock are useful in the same instance but what
>about between instances, my application have 10 instances.
>
You don't need to worry about this.
>I've a ILLRequestLock table used when I have to lock
>a ILLRequest row to allow only one writer and many readers.
>
Won't EOF's optimistic locking handle this?
>Is it required to lock/unlock editing context where all the entities
>are read-only in EOModeler?
>
Yes, unless you have them in teh EOSharedEditingContext.
Chuck
--
Chuck Hill email@hidden
Global Village Consulting Inc. http://www.global-village.net
_______________________________________________
webobjects-dev mailing list | email@hidden
Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/webobjects-dev
Do not post admin requests to the list. They will be ignored.