What means the message "potential deadlock"/"full thread dump will follow as debug output"

Hey there!

I just have a doubt and want to understand what happens. This isn't affecting the use of the tool, but if there's anything I could do to avoid/improve, it would be great.

I'm woking mainly with Text Analysis for social media, and eventually the dataset is pretty big (100K tweets, for example).

In those cases, especially, Knime seems to struggle a bit (it becomes irresponsive for some minutes). In this scenario, I got those messages:

WARN  KNIMEApplication$4              Potential deadlock in AWT Event Queue detected. Full thread dump will follow as debug ouput.
WARN  KNIMEApplication$3              Potential deadlock in SWT Display thread detected. Full thread dump will follow as debug ouput.

After the message, and after looking that I would need to restart Knime for a few minutes, Knime is back on action and completes the workflow.

As mentione, I'm just trying to understand what this message means, and if there's anything I could do to not cause it.

Thanks!

Gustavo Velho

Hi Gustavo,

you might want to look here: https://docs.oracle.com/cd/E19205-01/820-0619/geokj/index.html 

In short it is a Java Problem which is mainly based on the GUI.

I will ask our developers to look into it.

Best, Iris 

Thanks Iris!

Let me try to clarify what the "Potential deadlock in AWT Event Queue detected." message means. 

As we try to express in the message this is often only a potential deadlock. If it's a real deadlock is hard to determine programmatically but something is fishy if that message occurs. This could mean:

  1. A real deadlock / KNIME bug... in this scenario the application will be frozen and there is no chance to recover. You need to kill the application (I would wait much longer before actually doing this ... just to make sure it it's not (3)). I would then try to reproduce the problem and share it here with the us so that we can address it.
  2. A system deadlock - this used to happen sometimes on Linux/Mac in the past but thanks to a recent eclipse update this is much more seldom as it used to. You have to kill the application then.
  3. A temporary "hiccup" from which the application recovers within a short moment (few minutes max). This may happen if you, for instance pass in very wide data so that the metadata to the data is very complex (say, hundreds of thousands of columns).

Luckily (1) and (2) and very seldom (we are not aware of any concrete problem). In any case if you suspect a problem it's extremely helpful that you attach the knime.log file and possibly the workflow+data when you post it here. It allows us to diagnose the problem.

Hope this helps,
  Bernd