[infinispan-issues] [JBoss JIRA] (ISPN-8411) Add support for efficient removeAll

[Radim Vansa (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-8411?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13477977#comment-13477977 ] 

Radim Vansa commented on ISPN-8411:
-----------------------------------

[~papegaaij] ClearCommand does not participate in transaction; it is broadcast to other nodes and wipes out data without any specific ordering. It does not lock entries. When you do (in single transaction) cache.get(K); cache.clear(); cache.get(K); the second read will still return the original entry (if you use repeatable-read isolation).
The problem is that if the command were invoked transactionally, it would require locking whole cache, blocking all the other operations, which is not acceptable. In fact the {{Caches.removeAll}} in Hibernate does exactly this.

About the configuration - yes, 8.2 documentation was not updated at that time, 9.0 already points to Hibernate docs which is up-to-date, too, and explains the valid options. Hibernate ORM currently has non-tx caches as default (since 5.1 I think), WildFly is rather slow to adopt it.

It is not wrong (in Hibernate 5.x we wanted to support old configurations to allow for some bake-in and fallback), just a tad slower, and you may have issues like the one you experience. For Hibernate 6.0 the transactional caches will be likely discontinued as there are no known benefits in using them and it's a maintenance burden.

> Add support for efficient removeAll
> -----------------------------------
>
>                 Key: ISPN-8411
>                 URL: https://issues.jboss.org/browse/ISPN-8411
>             Project: Infinispan
>          Issue Type: Feature Request
>          Components: Hibernate Cache
>    Affects Versions: 8.2.8.Final, 9.1.1.Final
>         Environment: WildFly 10.1.0, WildFly 11.0.0.CR1, WildFly master, Hibernate 2LC
>            Reporter: Emond Papegaaij
>            Assignee: Galder Zamarreño
>
> Infinispan currently does not seem to implement an efficient way to clear an entire cache cluster-wide. This forces Hibernate to remove all entries one by one when a cache region needs to be cleared, for example when a buld CriteriaUpdate or CriteriaDelete is used.
> The behavior we are observing is:
> # All nodes in the cluster are queried for the keyset in a region
> # A lock seems to be in place for this region for the duration of the commit
> # The initiating node constructs a message with {{InvalidateCommands}} for all keys
> # This large message (230MB for 200k entries) is sent to all nodes in the cluster
> For large caches this can take very long. We had to increase the remote-timeout to 60 seconds to prevent timeouts. During this time, the entire cluster is locked an busy processing the cache invalidations. As you can understand, this is not a workable solution for us. On some places we can prevent the cache clear by updating the records one by one, but in other places this is not an option. 
> The corresponding report at Hibernate can be found here: https://hibernate.atlassian.net/browse/HHH-12036



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)