Re: Deadlock between spark logging and wildfly logging

2014-11-28 Thread Sean Owen
I'm kind of guessing here, but it looks like the log4j class has locked its own internal data structure (ConsoleAppender) and then has a lock on System.err. The JBoss logger looks like it swaps out System.err (?) with its logging stub, and has that locked, and then the stub routes to log4j to log,

Re: Deadlock between spark logging and wildfly logging

2014-11-28 Thread Charles
Here you go. "Result resolver thread-3" - Thread t@35654 java.lang.Thread.State: BLOCKED at java.io.PrintStream.flush(PrintStream.java:335) - waiting to lock <104f7200> (a java.io.PrintStream) owned by "null_Worker-1" t@1022 at org.jboss.stdio.StdioContext$DelegatingPri

Re: Deadlock between spark logging and wildfly logging

2014-11-28 Thread Sean Owen
Are you sure it's deadlock? print the thread dump (from kill -QUIT) of the thread(s) that are deadlocked, I suppose, to show where the issue is. It seems unlikely that a logging thread would be holding locks that the app uses. On Fri, Nov 28, 2014 at 4:01 PM, Charles wrote: > We create spark cont