Out of memory error when running workflows

Hi,
We have a Knime server (version 3.6.2) that is running on Ubuntu-16.04. Whenever we try to run a workflow from the server we get the next error:
image

We have tried to shutdown/restart the server but to no avail.

Hi @RaulPerez,

could you please quick check which java version is actual installed on the server by running the following command?

java -version

it might be that you have stumbled over the java issue here: Java update 8u242 breaks execution with RMI executors

I assume that you are using either KNIME server version 4.7 or 4.8 with executor version 3.6.2? Than I would suggest to update your server to a more recent version like 4.10. If you are already on version 4.8 you can also update the server to 4.8.4.

All downloads and update guides are linked here: https://www.knime.com/knimecom-product-downloads

Best regards,
Michael

1 Like

Hi @MichaelRespondek,

Thank you for the advice. This is the Java version we are running:
image

We are currently trying to update our server version.

My IT teams is asking if we could call you to speed things up (using zoom or GoogleMeet). We know it isnt “normal” but we would greatly appreciate the help.

Hi @MichaelRespondek

My IT team sends this additional information:

HI @RaulPerez,

sorry, I only can help you with the update by giving a quick overview how I would do this:

As your installation is some versions behind the recent I would install the 4.10.3 using the installer beside the existing installation. If you have a internet connection the executor can be installed also within the server installation. Otherwise you can download the executor build here: https://www.knime.com/knimecom-product-downloads-2019-12

It comes with many extensions installed already (doesn’t depend on the way it is installed (server installer vs. separate download).

After the installation stop the started server and merge your configurations of the old install to the new config files, referenced here:

https://docs.knime.com/2019-12/server_admin_guide/index.html#server-configuration-files-and-options

Then you have to copy the workflow repository of the old server installation to the new.

Now you can start the fresh installed server, the installation as a service is described here:

https://docs.knime.com/2019-12/server_installation_guide/index.html#_defining_a_linux_runlevel_script

Hope this helps you updating the server.

Update guide for reference: https://docs.knime.com/2019-12/server_update_guide/index.html

Btw. your OpenJDK version indicates that you are affected by the linked issue in my first reply.

Cheers,
Michael

1 Like

Hi @MichaelRespondek

Thank you for your answer. We understand.

My IT team would like to know if you could recommend us a size for our aws machine? They comment that even with more memory and the CPU it has right now it gets overloaded.

Hi @RaulPerez,

sorry for the late reply. Unfortunately we can’t give recommendations on aws machines as the use cases for the KNIME server are so various. But as there are many different aws types you should go for one with many cores and a big amount of RAM.

If you have the executor rotation turned off (default on 4.10.x) you can assign ~70 % of the available RAM to the executor. https://docs.knime.com/2019-12/server_admin_guide/index.html#knime.ini-file
If your server has to handle many jobs in parallel you could consider to assign more RAM to the server process.

https://docs.knime.com/2019-12/server_admin_guide/index.html#server-startup-takes-a-long-time

Both options are configured via the -Xmx parameter.

1 Like