tristantarrant commented on issue #3399: URL: https://github.com/apache/logging-log4j2/issues/3399#issuecomment-2622169191
Neither of those suggestions improved things. However I have some additional info. I've put some good old `System.out.printf`s around the write lock code. Here is an example of things going well: ``` writeLock - ask - org.infinispan.notifications.cachemanagerlistener.CacheManagerNotifierImpl, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d writeLock - acquire - org.infinispan.notifications.cachemanagerlistener.CacheManagerNotifierImpl, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d writeLock - ret - org.infinispan.notifications.cachemanagerlistener.CacheManagerNotifierImpl, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d / c=org.infinispan.notifications.cachemanagerlistener.CacheManagerNotifierImpl:DEBUG in Default n=org.infinispan.notifications.cachemanagerlistener.CacheManagerNotifierImpl:DEBUG in Default writeLock - release - org.infinispan.notifications.cachemanagerlistener.CacheManagerNotifierImpl, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d ``` the `ret` part is what is being returned: `c` is the `currentLogger` and `n` is the `newLogger`. Here is an example when bad things start happening. I get 100s of the following: ``` writeLock - ask - org.infinispan.LIFECYCLE, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d writeLock - acquire - org.infinispan.LIFECYCLE, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d writeLock - ret - org.infinispan.LIFECYCLE, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d / c=null n=org.infinispan.LIFECYCLE:DEBUG in Default writeLock - release - org.infinispan.LIFECYCLE, org.apache.logging.log4j.message.ReusableMessageFactory@6dab9b6d ``` The Weak semantics are causing it to not find the entry in the map, so every single request for the `org.infinispan.LIFECYCLE` acquires the write lock, skipping the "fast path". Changing the `loggerRefByNameByMessageFactory` to use a normal HashMap and removing the WeakReferences fixes the issue. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: notifications-unsubscr...@logging.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
