[infinispan-issues] [JBoss JIRA] (ISPN-10171) Spring Cache sync locking does not work

[Katia Aresti (Jira)]

     [ https://issues.jboss.org/browse/ISPN-10171?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Katia Aresti updated ISPN-10171:
--------------------------------
    Git Pull Request: https://github.com/infinispan/infinispan/pull/6897, https://github.com/infinispan/infinispan/pull/6921  (was: https://github.com/infinispan/infinispan/pull/6897)


> Spring Cache sync locking does not work
> ---------------------------------------
>
>                 Key: ISPN-10171
>                 URL: https://issues.jboss.org/browse/ISPN-10171
>             Project: Infinispan
>          Issue Type: Bug
>          Components: Spring Integration
>    Affects Versions: 9.4.11.Final, 10.0.0.Beta4
>         Environment: > mvn -version
> Apache Maven 3.6.0 (97c98ec64a1fdfee7767ce5ffb20918da4f719f3; 2018-10-24T20:41:47+02:00)
> Maven home: /usr/local/Cellar/maven/3.6.0/libexec
> Java version: 11.0.1, vendor: Oracle Corporation, runtime: /Library/Java/JavaVirtualMachines/openjdk-11.0.1.jdk/Contents/Home
> Default locale: en_NL, platform encoding: UTF-8
> OS name: "mac os x", version: "10.14.4", arch: "x86_64", family: "mac"
>            Reporter: Rens van Leeuwen
>            Assignee: Katia Aresti
>            Priority: Major
>
> With ISPN-7224/https://github.com/infinispan/infinispan/pull/5262, an implementation was written to support a synchronous get that implements Spring Cache's {{@Cacheable(sync = true)}}.
> That seems to work well in isolation, but not in an integrated way. By that I mean, that the implementation of {{SpringCache}} maintains its own {{ReentrantLock}}s in each instance.
> Zooming out a bit, this is part of the call stack when the {{SpringCache}} instance is looked up during handling the {{@Cacheable(sync = true)}} annotation:
> {noformat}
> CacheInterceptor.invoke(MethodInvocation)  (org.springframework.cache.interceptor)
>   CacheAspectSupport.execute(CacheOperationInvoker, Object, Method, Object[])  (org.springframework.cache.interceptor)
>     CacheOperationContexts in CacheAspectSupport.CacheOperationContexts(Collection<? extends CacheOperation>, Method, Object[], Object, Class<?>)  (org.springframework.cache.interceptor)
>       CacheAspectSupport.getOperationContext(CacheOperation, Method, Object[], Object, Class<?>)  (org.springframework.cache.interceptor)
>         CacheOperationContext in CacheAspectSupport.CacheOperationContext(CacheOperationMetadata, Object[], Object)  (org.springframework.cache.interceptor)
>           CacheAspectSupport.getCaches(CacheOperationInvocationContext<CacheOperation>, CacheResolver)  (org.springframework.cache.interceptor)
>             AbstractCacheResolver.resolveCaches(CacheOperationInvocationContext<?>)  (org.springframework.cache.interceptor)
>    ----->>    SpringEmbeddedCacheManager.getCache(String)  (org.infinispan.spring.embedded.provider)
>                 SpringCache.SpringCache(BasicCache)  (org.infinispan.spring.common.provider)
>                   SpringCache.SpringCache(BasicCache, long, long) (org.infinispan.spring.common.provider)
> {noformat}
> The problem with this is, that for each turn the interceptor handles the request a *new instance* of the {{SpringCache}} is made in the {{SpringEmbeddedCacheManager}}, rendering the use of the {{ReentrantLock}} mechanism useless to a certain extent, because multiple threads accessing the {{@Cacheable(sync = true)}} method will each have their own {{SpringCache}} and corresponding {{ReentrantLock}}.
> In turn, there is no locking at all, and each thread will be executing the (assumed to be) expensive code that is intended to be cached.
> --
> I'm intending to write a patch for this, but this would be the first time I commit to an open source project (bear with me ;)). I'll try to provide a pull request for this.



--
This message was sent by Atlassian Jira
(v7.12.1#712002)