Weblogic related gotchas / useful config options

Weblogic Configs

1. On Linux / WLS 12c there is a problem with the TCP unicast work manager blocking, it causes nodes in a cluster to failover. Add:
    
    -Dweblogic.unicast.HttpPing=true

To server start args. It's a known / documented issue: https://docs.oracle.com/cd/E24329_01/doc.1211/e26593/issues.htm#WLSRN182

2. Memory leaks. If you use RMI which you probably do and suffer from Memory leaks, forget tuning the app code (it's Java,
it'll always have memory leaks) Take a heap dump adn run it through jmap or create diffs from 2. If you see this class pop
up in retained heap or memory leak suspects: weblogic.rjvm.ConnectionManage Add:

    -Dweblogic.system.StreamPoolSize=0
    
To server start args. It's not documented anywhere but from Oracle Support: 

=================================================================================================
Analyzed the heap dump and found the below Problem suspect: 

2,933 instances of "weblogic.rjvm.ConnectionManagerServer", loaded by "sun.misc.Launcher$AppClassLoader @ 0x6c00075d8" occupy 
1,942,365,888 (77.22%) bytes. These instances are referenced from one instance of 
"weblogic.utils.collections.ConcurrentWeakHashMap$Segment[]", loaded by "sun.misc.Launcher$AppClassLoader @ 0x6c00075d8" 

In my research I found that weblogic.rjvm.ConnectionManager will be used for managing a pool of steams for replication/serialization ....
as it uses a pool it can be considered as leak because streams are not dropped after the object serialization. 

The property weblogic.system.StreamPoolSize when set to 0 disable the cache pooling, of course that will consume less memory in
particular if big objects to serialize. 

Please set the property :-Dweblogic.system.StreamPoolSize=0 and test again.
=================================================================================================

Pre -Dweblogic.system.StreamPoolSize=0 argument, GC recovering less and less each cycle:

Nightly restarts had been accepted as a way of life (sigh..), -Dweblogic.system.StreamPoolSize=0 argument put live on 21/10. GC less often and reclaims more heap each cycle! Restarts are not a solution!

Thanks a ton. We are also facing the same problem in production and we would like to give a try with -D arg. We have not seen any reference for this other than this blog.

Missed this comment @kotari4u. How did it work out?