Why not instantiate a ServletContextSymbolProvider (the ServletContext is available from the
ApplicationGlobals service) and contribute it to the SymbolProvider before ApplicationDefaults and
FactoryDefaults. That way you can directly override tapestry.page-pool.hard-limit and automatically
have all other init parameters as symbols, too.
Uli
9902468 schrieb:
I always keep nabble open when I'm at work :)
These parameters can be transferred to web.xml, so that recompile is not
needed when changing these parameters:
public static void contributeApplicationDefaults(MappedConfiguration<String,
String> configuration, ApplicationGlobals globals) {
configuration.add("tapestry.page-pool.hard-limit",
globals.getServletContext().getInitParameter("page_hard_limit"));
}
and in web.xml
<web-app>
<display-name>Some Tapestry 5 Application</display-name>
<context-param>
<!-- The only significant configuration for Tapestry 5, this informs
Tapestry of where to look for pages, components and mixins. -->
<param-name>tapestry.app-package</param-name>
<param-value>fi.hanken.resbuilder.tapestry5</param-value>
</context-param>
<context-param>
<param-name>page_hard_limit</param-name>
<param-value>60</param-value>
</context-param>
This can be used through out every single service and page and component. To
access this in a page or component use:
@Property @Inject @Value("${page_hard_limit}")
private String pageHardLimit;
Injection can be used in service constructor also (I think) or by using
buildSomeService method instead of the usual binder.bind
Of course T5 makes it easy to read these from a file or database or from
what not, but I have noticed that many admins accept editing web.xml and
restarting the app ok.
Hopefully this helps. (And makes it easier to say to admins that just
configure the app according to the load they are experiencing.)
-99
nille hammer wrote:
Man that was a quick answer! Thanks for that. I am aware of this options
in the configuration. I had not tuned them before, because load is not
my problem. Anyway I have tuned them now and am currently waiting for
the admistrator to deploy the newest war.
According to acces-logs I was the only user accessing the page. So as
far as I understand the mechanism only one instance of index should have
been created and not 20. I think it is an issue with configuration in
production environment. But I have no idea where to take a closer look
(i.e. where to build in log-messages, which of Tomcat's configuration
options to check etc.).
So suggestions are still more than welcome.
Nillehammer
9902468 schrieb:
Hi,
Eventually we had to add these to appmodule:
configuration.add("tapestry.page-pool.hard-limit", "60");
configuration.add("tapestry.page-pool.soft-limit", "20");
configuration.add("tapestry.page-pool.soft-wait", "20 ms");
as instructed here:
http://tapestry.apache.org/tapestry5/guide/conf.html
This solved our problems, and you can go up as long as your server has
memory. Our problems were only in production, as development stress just
isn't enough. (This is why one really should do load testing with jmeter
or
similar tool.)
The defaults could be a bit more? Or considerably less, so that the
developer would hit these walls during development and understands to
load
test / tune the parameters.
- 99
nille hammer wrote:
Hi list,
I am also facing the problem that 9902468 had. I have -according to
Howard's suggestion- replaced every pageLink with a customPageLink to
prevent targeted pages from being created unecessarily. The problem
persists. Trying to load the index-Page of my application with a browser
I have to wait about two minutes before I get an empty page. In the
app's log the following Exception appears:
java.lang.RuntimeException: The page pool for page 'Index' (in locale
de) has been exausted: there are 20 instances currently being used and
no more can be created. Try increasing the hard limit (symbol
tapestry.page-pool.hard-limit) to allow additional instances to be
created, or increasing the soft wait (symbol
tapestry.page-pool.soft-wait) to trade away some throughput for more
efficient use of page instances.
at
org.apache.tapestry5.internal.services.PagePoolCache.checkout(PagePoolCache.java:199)
at
org.apache.tapestry5.internal.services.PagePoolImpl.checkout(PagePoolImpl.java:107)
at $PagePool_11c76f8cf9a.checkout($PagePool_11c76f8cf9a.java)
at
org.apache.tapestry5.internal.services.RequestPageCacheImpl.get(RequestPageCacheImpl.java:43)
at
$RequestPageCache_11c76f8cf99.get($RequestPageCache_11c76f8cf99.java)
at
$RequestPageCache_11c76f8cf91.get($RequestPageCache_11c76f8cf91.java)
...
In development and in production environment I use Java 6 and Tomcat 6.
Tapestry is 5.0.14. The strange thing is that I am only seeing this in
production environment.
Has anyone a suggestion how to track this down? I am pretty stuck with
this.
Thanks in advance, nillehammer
Howard Lewis Ship wrote
Tapestry is obtaining page instances to see if there is an activation
context for the page.
I expect to optimize this in the future, to check to see if the page
has an event handler for the passivate event.
In the meantime, you could build your own PageLink component that
generates the link by invoking
ComponentResources.createPageLink("pagename", true) ... that will tell
Tapestry to generate a link to the page with no activation context,
and the page will NOT be queried for its context, and therefore, won't
need to be attached to the current request.
On Fri, Aug 29, 2008 at 3:58 AM, 9902468 <[EMAIL PROTECTED]>
wrote:
Hi,
we are experiencing strange hard pool limit reached exceptions. (For
the
user count that we have.) After debugging a little, it seems that our
left
menu component, which is full of page links, instantiates not only the
page
that we are at, but also the ones that are referenced in page link.
This can be seen also in that exception page is shown, if one of the
pages
that are referenced in page links have an error.
Is this correct behavior? And should we really use common a href= tags
to
avoid this? Further more, is it possible to create infinite loop, when
page
a has page link to page b, which has page link to page a?
- 99
--
View this message in context:
http://www.nabble.com/Page-pool-hard-limit-and-page-instance-usage-tp19217773p19217773.html
Sent from the Tapestry - User mailing list archive at Nabble.com.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
