I have one problem with performance in production because of, as I think, blocked threads .

When I got thread dumps from production environment I was surprised that about 50% of thread are in status:

-- Blocked trying to get lock: ch/qos/logback/classic/sift/SiftingAppender@0x1111fe3e8[unlocked]

At the same time, as you can see, SiftingAppender is 'unlocked'.

I have no idea why it's happened.

If we deep into logback sources we will find there a synchronized method which is the 'thin' place I think:

ch.qos.logback.core.AppenderBase.doAppend(E)

public synchronized void doAppend(E eventObject) {
    // WARNING: The guard check MUST be the first statement in the
    // doAppend() method.

    // prevent re-entry.
    if (guard) {
      return;
    }
    ...
  }

We use weblogic v.10.3.5.0

slf4j.version - 1.7.5

logback.version - 1.0.13

Has anybody ideas why it's happened and how to avoid this problem?

It may be not logback's problem. I now trying to find resolution only.