[hibernate-issues] [Hibernate-JIRA] Commented: (HHH-7256) Infinispan second level cache should react to minimal puts in putFromLoad impl

[Steve Ebersole (JIRA)]

    [ https://hibernate.onjira.com/browse/HHH-7256?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=46327#comment-46327 ] 

Steve Ebersole commented on HHH-7256:
-------------------------------------

Moving updating of stats to the cache providers is not a good solution.  First you are duplicating that code, potentially fragmenting it.  Second, stats is a critical thing relied on by many applications; this suggestion makes them dependent that integration code updates stats properly.

The reasoning you gave was because you thought Hibernate was "guessing" at what the cache provider is doing internally.  But my point is that we are not guessing.  All put methods expect the cache provider to return whether or not a put actually happens as well as an argument indicating "minimal puts setting" for the current call context.  In the particular case we are discussing, the Infinispan integration is just returning the wrong value.

> Infinispan second level cache should react to minimal puts in putFromLoad impl
> ------------------------------------------------------------------------------
>
>                 Key: HHH-7256
>                 URL: https://hibernate.onjira.com/browse/HHH-7256
>             Project: Hibernate ORM
>          Issue Type: Improvement
>          Components: caching (L2)
>            Reporter: Galder Zamarreno
>            Assignee: Galder Zamarreno
>             Fix For: 4.1.3
>
>
> When entities are loaded multiple times, it could happen that two consecutive Cache.putForExternalRead() operations are called with the second not affecting the cache if a an entry is already present.
> However, from a cache put statistics will increase, resulting potentially in a huge number when in reality, the number of actual cache updates has been pretty small. The minimalPuts parameter is there to avoid this, implying that maybe a check if the key is present should be added since PFER does not return a boolean or similar.
> Although adding that check looks something obvious, it's something that we're probably reluctant to do, because since caches are configured with storeAsBinary=true, it could lead to unmarshalling, slowing things down.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira