KNIME 2.9.0 hangs under Centos 6.5 (sometimes)

Apologies in advance if this isn't the correct forum.

I'm seeing hangs in KNIME 2.9.0 when a user is editing their workflows (the environment is Centos 6.5 x86_64). The hangs occur sporadically and cause the entire KNIME interface to freeze (none of the GUI elements respond) and CPU use to go to 100% of 1 core.

When I examine the JVM using jstack and jvisualvm, the AWT-XAWT thread is consuming 100% of the runtime.

Unfortunately, I can't replicate the issue on another machine, but I can't think what might be special about the one that is showing the problem.

Is this a known issue? Can anyone suggest a workaround or what might be causing this problem?

I've attached the jstack output and the KNIME configuration from Help->"About KNIME"->"Installation Details" panel.

Hi Jonathan, 

Not sure, but looking into it.  Sorry for the hassle.

Regards,

Aaron

Hi Jonathan,

Thanks for the problem report. This has also come up at another place and we spent some time to reproduce it ... successfully. We could also isolate the problem in a non-KNIME environment and have searched the bug databases at java and eclipse. 

We believe this Java bug to be the culprit. It has been fixed in "JRE7u60", which is currently available as early access release (current latest is JRE7u51 and we are bundling JRE7u07). According to our tests the 'u60' version indeed fixes the issue. However, as it hasn't been officially released and - as you say - the problem only occurs sporadically we will not release a hot fix.

(If you want to test it: Download the u60 version and replace the "jre" directory in your KNIME installation folder with the new JRE. Please keep in mind that the early access release is not considered to be stable and could cause different problems.... which you can of course share!)

Regards,
Patrick

A side note... You can keep the old jre and start KNIME with the new one, by changing the knime.ini file. Just put

-vm
/path/to/your/jre

before the -vmargs line.

Cheers, gabor

 

The change in the JVM seems to have fixed it for my users - although it's early to say definitively.

Many thanks for everyone's help.