Re: 5.3 CLOSE_WAIT problem and multi instance change notification
Re: 5.3 CLOSE_WAIT problem and multi instance change notification
- Subject: Re: 5.3 CLOSE_WAIT problem and multi instance change notification
- From: Hordur Thordarson <email@hidden>
- Date: Wed, 12 Oct 2005 18:18:37 +0000
On 12.10.2005, at 17:30, Chuck Hill wrote:
Just a few more questions:
1) At the risk of exposing my ignorance, how do I get a thread dump
from a running java app, preferably via Terminal ?
Well, if you had the book you could start reading on page 92. :-)
Ok, I get the message now :-)
If they don't communicate the change with change notification then
the only ways I can see are the ones suggested by you/Apple:
- Timed invalidation via EO(F)'s fetch timestamps
This applies only to the first fetch/fault of an object into an
editing context or subsequent refaulting of particular objects.
Hmm, sounds like I need to re-read this. I was assuming that
functioned like so given a 60 minute refetch timeout:
- An app requests an EO for the first time and it gets fetched from the
db and gets the current timestamp
- 30 minutes later the app re-requests the EO and gets the cached
version from the object graph
- 2 hrs down the road the app re-requests again and EOF then goes to
the db again as the EO is over 60 minutes old in the cache
Am I misunderstanding this ?
My test case is a CMS with a Direct Action based front end (display)
and a Component based back end (editing). If I'm running two
instances of the app and I open up two browser windows, one
displaying a page from the CMS and the other for editing that page,
then every time I refresh the DA page I'll switch app instances due
to round robin load balancing, so sometimes I'll be looking at data
from the same instance I'm editing the page in which is easy to keep
in sync, but sometimes I'll be looking at data from the other app
instance, which is harder to keep fresh.
This scenario must occur fairly often in WO apps that allow editing
of their and it would be great to hear how people are dealing with
this.
Someone else (Ken?) already gave you my answer for this: a preview
URL. Keep the editor and preview in the same session so that they see
the same data. For regular viewers, they can get fresh data when
their session is created / when they first look at something. After
that, they see what they originally saw, the information does not
change on them as they browse. That may or may not work for you.
Uti had some suggestioins and it sounds like the only way of doing this
is to make sure the preview url contains the same app number and
session id as the user's editing session. Unfortunately users are not
the easiest bunch of people to control so inevitably they will first
load a web page through the front end, and then go edit it in another
window in the backend (and possibly in another app instance), and then
they will scream when they refresh the original page and their changes
don't show up :-)
Anyway, thanks for the help !
Hordur
On 11.10.2005, at 22:57, Chuck Hill wrote:
Hi,
On Oct 11, 2005, at 11:52 AM, Hordur Thordarson wrote:
Hi all,
I have a client who is running an app of mine on OSX Server 10.4
with WebObjects 5.3, and recently we've been having significant
problems with the app, probably as the load on the app has been
increasing.
Currently there is just 1 instance of the application running. It
will run fine for a while after being restarted, but then the app
will go into a state where according to top it has > 100 threads as
opposed to around 40-70 normally, and it will become unresponsive.
Doing a netstat at this point in time shows a lot of connections in
a CLOSE_WAIT state and the only way out is to kill the instance and
restart the WebObjects service.
Is anyone seeing anything like this with OSX Server 10.4 and WO 5.3
?
Not normally, but your description of > 100 threads sounds like it
is deadlocked somewhere. Have you tried getting a thread dump?
Also, I was wondering if maybe running multiple instances would
help with this
Even having 40 - 70 threads in one instance says that you want to be
running multiple instances. That many threads indicates that
something is blocking / deadlocking or that that instance is way
overloaded.
and wanted to ask what people here are using for multi-instance db
change notification. I've read about Project Wonder's
ERChangeNotificationJMS and there is some old code from David
Neumann of Apple floating around but that's all I've been able to
find. Are you guys using ERChangeNotificationJMS or are you using
homegrown stuff for multi instance sync ?
ERChangeNotificationJMS is a modern version of David Neumann's code.
If I was going to use such a notification system that is the one
that I would choose. So far, I have not found it to be necessary.
You need to plan for object freshness and create new editing
contexts / do refreshing fetches to ensure that you have data as
fresh as you need it. You also need to take care of optimistic
locking conflicts, but those can also happen when using a change
notification system.
Chuck
--
Coming in 2006 - an introduction to web applications using
WebObjects and Xcode http://www.global-village.net/wointro
Practical WebObjects - for developers who want to increase their
overall knowledge of WebObjects or who are trying to solve specific
problems.
http://www.global-village.net/products/practical_webobjects
--
Coming in 2006 - an introduction to web applications using WebObjects
and Xcode http://www.global-village.net/wointro
Practical WebObjects - for developers who want to increase their
overall knowledge of WebObjects or who are trying to solve specific
problems.
http://www.global-village.net/products/practical_webobjects
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden