[infinispan-issues] [JBoss JIRA] (ISPN-7105) Expiration is not working under some circumstances with AtomicMap

Dan Berindei (JIRA) issues at jboss.org
Thu Apr 27 06:10:00 EDT 2017


     [ https://issues.jboss.org/browse/ISPN-7105?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dan Berindei resolved ISPN-7105.
--------------------------------
    Fix Version/s: 8.1.0.Final
       Resolution: Done


The problem could be reproduced with regular put operations as well, it only requires a read operation on the same key in the same transaction (and atomic maps always try to read an existing map entry before creating a new one). 

The read created a {{ClusteredRepeatableReadEntry}} in the context, with a {{SimpleClusteredVersion.NON_EXISTENT}} version and no lifespan. The write then tried to update the context entry's metadata with the default lifespan in {{EntryFactoryImpl.wrapEntryForPut()}}, but {{EntryFactoryImpl.updateMetadata()}} ignored the update.

The problem was fixed with ISPN-5699 (included in version 8.1.0.Final).



> Expiration is not working under some circumstances with AtomicMap
> -----------------------------------------------------------------
>
>                 Key: ISPN-7105
>                 URL: https://issues.jboss.org/browse/ISPN-7105
>             Project: Infinispan
>          Issue Type: Bug
>          Components: Core
>         Environment: AtomicMap, write-skew, transactional 
>            Reporter: Wolf-Dieter Fink
>            Assignee: Dan Berindei
>             Fix For: 8.1.0.Final
>
>         Attachments: jdg-ws-am-lifespan-reproducer.zip
>
>
> When AtomicMaps are used with writeskew (+ simple versioning) lifespan from the config file is ignored and instead get infinite lifespan for the entries.
> The configured default value is used if I any of the following parameters change
> - Disable write skew checks
> - Use a normal put (not atomic maps)
> - Do not start transactions first.



--
This message was sent by Atlassian JIRA
(v7.2.3#72005)


More information about the infinispan-issues mailing list