Consultation on KNIME Business Hub (4 vCore / 32 GB) Capacity Planning and Concurrency Recommend...

Hi Team,

We would like to seek your advice regarding the capacity planning and concurrency capability of a KNIME Business Hub environment.

Current Environment

• KNIME Business Hub License: 4 vCore

• Memory: 32 GB RAM

• Deployment: Kubernetes

Workflow Characteristics

Our use case relies heavily on parent workflows calling sub-workflows via the Call Workflow node:

• Parent workflows invoke sub-workflows using Call Workflow nodes.

• Each sub-workflow contains approximately 200 nodes.

• Multiple workflows may run concurrently.

Observations During Load Testing

During a recent load test, we observed the following:

• CPU utilization reached nearly 100%.

• All 4 vCore tokens were fully consumed.

• Some workflows failed within a few seconds after starting.

• The Inspect function occasionally became unavailable.

• The following error appeared in the logs:

copy to log: unexpected EOF

In addition, we have attached the error log files generated during the load test for your analysis. Unfortunately, we do not have a complete Support Bundle and only retained the error logs.

Questions

1. Based on the workload described above, is a 4 vCore / 32 GB environment sufficient?

2. For a scenario where:

   • Call Workflow is heavily used; and

   • each sub-workflow contains approximately 200 nodes,

   does KNIME have any recommendations regarding the maximum number of concurrently running workflows?

3. Does KNIME provide any best practices or sizing recommendations regarding:

   • CPU sizing;

   • memory sizing;

   • Executor scaling;

   • recommended concurrent workflow limits?

4. Could the workflow failures and the copy to log: unexpected EOF errors observed during the load test be related to insufficient resources, particularly an insufficient number of vCores, or other architectural limitations?

If it is not possible to provide an exact number of concurrent workflows, we would greatly appreciate an estimated or recommended concurrency range, or a methodology for capacity planning, so that we can properly size the environment.

Any guidance regarding sizing recommendations and best practices would be highly appreciated.

Thank you very much for your support.