[
https://issues.apache.org/jira/browse/HADOOP-11771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14385165#comment-14385165
]
Arun Suresh commented on HADOOP-11771:
--------------------------------------
[~gopalv], looking at the code, It looks like there is a reason for
CACHE_CLASSES to be a WeakHashMap (to be able to drop the cached classes when
the classloader goes away) and there is a reason why we need it synchronized.
Id say you could fix this using a {{ConcurrentWeakHashMap}} which we would have
to probably write from scratch (might not be too hard, but I dont expect it to
be trivial). Also you have to note that the actual concurrency level (atleast
in the case of {{ConcurrentHashMap}}) has to be specified in the constructor
when it is created, else it will be a default value for the entirety of its
lifetime.
Alternatively, is it possible to reuse a value obtained from an earlier call to
this method by the client code/framework/utility (which based on your stack
trace is the hive planner)?
> Configuration::getClassByNameOrNull synchronizes on a static object
> -------------------------------------------------------------------
>
> Key: HADOOP-11771
> URL: https://issues.apache.org/jira/browse/HADOOP-11771
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: conf, io, ipc
> Reporter: Gopal V
> Attachments: configuration-cache-bt.png, configuration-sync-cache.png
>
>
> {code}
> private static final Map<ClassLoader, Map<String, WeakReference<Class<?>>>>
> CACHE_CLASSES = new WeakHashMap<ClassLoader, Map<String,
> WeakReference<Class<?>>>>();
> ...
> synchronized (CACHE_CLASSES) {
> map = CACHE_CLASSES.get(classLoader);
> if (map == null) {
> map = Collections.synchronizedMap(
> new WeakHashMap<String, WeakReference<Class<?>>>());
> CACHE_CLASSES.put(classLoader, map);
> }
> }
> {code}
> !configuration-sync-cache.png!
> !configuration-cache-bt.png!
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
