Thanks for all your help and your patience Jan. With this last bit of information we are ready to try out the caching in test, staging en production environments. The picture of caching for multisite is now complete for me. Whenever I find time (will not be sooner than september I'm afraid) I can put all necessary steps in a wiki page for everyones benefit.
Bye, Bert From: [email protected] [mailto:[email protected]] On Behalf Of Jan Haderka Sent: maandag 2 augustus 2010 8:48 To: Magnolia User-List Subject: Re: [magnolia-user] cache filter configuration bypass does not work On Jul 30, 2010, at 10:11 AM, Bert Leunis wrote: Hello all, After more tests I can refine my question about the uuid-key-mapping cache. I built and used the FlushSubpathCacheListeningPolicy as shown in Jans blog (http://weblogs.java.net/blog/rah003/archive/2010/03/24/flush-cache). It works very well. I have these caches now: uuid-key-mapping default siteA siteB When I activate a page for siteA, three caches are flushed: siteA, default, uuid-key-mapping. The cache for siteB is untouched, what is exactly what I want. I think that the flushing of the uuid-key-mapping cache is unwanted though. There may be uuid's stored that reference elements of the cache of siteB. If you now lookup a uuid belonging to siteB in the uuid-key-mapping cache, because you want to flush all elements belonging to that uuid, you will not find it. And the element for that uuid remains in the cache of siteB. I hope I describe this situation correctly. Do you agree with me on this problem? I think not flushing the uuid-key-mapping cache is the solution here. There may stay extra keys in there, that reference elements that are already flushed, but that seems not such a big problem to me. Yes you are right, the uuid-key-mapping should not be touched. It's better to have obsolete keys then the missing ones. To fix that you can just remove all the repositories from the .../uuid-key-mapping/flushPolicy/policies/flushAll/repositories > Yeah, bypass should work as long as all the images are in that same place. Wrestling with the voters, but they all voted against me, I had no luck there. probably case of reversing some of the voters or increasing the "level" so it can override the others (like the site voter) But adapting the site-configuration did the job! Images with a url starting with "/.imaging/stk/hollanddoc" are now recognized as being part of the hollanddoc site by the SiteURIPatternVoter. In /modules/extended-templating-kit/config/sites/hollanddoc/mappings: - imaging +-- URIPrefix = /.imaging/stk/hollanddoc +-- handlePrefix = /.imaging/stk/hollanddoc +-- repository = imaging Cool. Glad you found a working solution. Now I am trying to grasp the purpose of the uuid-key-mapping cache. I think I understand what you wrote in your blog about that (http://weblogs.java.net/blog/rah003/archive/2010/03/10/cached-again). When an item with a uuid X is changed, using the uuid-key-mapping cache you can find all items in the default and other caches that contain cached content regarding uuid X. All these elements should be flushed. Is an implementation of this mechanism already used? Yes, this mechanism is used for flushing pages with enabled commenting. When the comment is made on the page, this is public-only change and this mechanism is used to find and flush only the pages associated with the comments. I hope you have some time to answer this question too. As always... much obliged! sorry for letting you wait, I was off for last couple of days ... Jan Thanks, Bert ________________________________ ---------------------------------------------------------------- For list details see http://www.magnolia-cms.com/home/community/mailing-lists.html To unsubscribe, E-mail to: <[email protected]> ---------------------------------------------------------------- ---------------------------------------------------------------- For list details see http://www.magnolia-cms.com/home/community/mailing-lists.html To unsubscribe, E-mail to: <[email protected]> ----------------------------------------------------------------
