You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@felix.apache.org by "Ioannis Canellos (JIRA)" <ji...@apache.org> on 2014/09/02 11:24:21 UTC
[jira] [Commented] (FELIX-4566) Consistency in PersistenceManager
and Cache is not guaranteed
[ https://issues.apache.org/jira/browse/FELIX-4566?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14118056#comment-14118056 ]
Ioannis Canellos commented on FELIX-4566:
-----------------------------------------
Even though getDictionaries() doesn't affect the consistency between cache/file persistance it should still be guarded so that it always returns the "snapshot" of the cache.
Regarding, the eager loading I am not sure. Personally, I think that since the service is backed by files its better to have an error during initialisation than later in the processes. For what it worths, I switched to eager loading so that I can be able to use a lock per pid. But since the getDictionaries() will require a global lock, then yes most probably will can switch back to lazy loading too.
> Consistency in PersistenceManager and Cache is not guaranteed
> -------------------------------------------------------------
>
> Key: FELIX-4566
> URL: https://issues.apache.org/jira/browse/FELIX-4566
> Project: Felix
> Issue Type: Bug
> Components: Configuration Admin
> Affects Versions: configadmin-1.8.2
> Reporter: Ioannis Canellos
> Assignee: Guillaume Nodet
> Attachments: FELIX-4566-2.patch, FELIX-4566.patch
>
>
> Currently the store method in the CachingPersistenceManagerProxy performs an update on the PersistenceManager and then also updates the Cache.
> Since we are updating 2 resources, without any form of synchronisation its possible that the resources are out of sync.
> For example:
> Two threads A and B call configuration.update() on the same pid. The first threads calls store(), the PersistenceManager gets updated and then the second thread kicks in updates both PersistenceManager and the cache and finally the first thread updates the cache.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)