Saving a workflow with 1 node takes more than 1 minute (thread dumps attached)

KNIME Analytics Platform
1

Dear Knimeler,
running my KNIME WF on a VM - details see below - I encountered the problem, that storing my WFs took very long. So I created a small example, consisting of 1 CSVReader node.

  1. Read a CSV File having 6mill rows/lines (9 columns; ~500mb)
  2. After having read the file, store the WF via menu icon
  3. => This storing took more than 1 minute on a VM having 320GB MEM, 40 cores, and SSD

Doing the same example on my laptop (SSD, 32GB, 8cores) takes 1-2 seconds. I added two thread-dumps (zipped)- perhaps you can see the reason.

Best
Erich
tdumps.zip (11.5 KB)

The Environment I was using:

  • VMware 6.0.0 build 3620759
  • 4 socket each with 10 cores
  • 320 GB RAM
  • Windows Server 2012 R2
    Version 6.3.9600
2

We experienced a similar problem recently on a Windows virtual machine. And it was the virus scanner. It seems it was configured to be very aggressive and each time KNIME would make a ‘move’ there was heavy scanning involved and also the start of KNIME was very slow (KNIME can consist of a few thousand files).

The solution was to exclude the KNIME directories from the scanner and schedule a scan instead at times when there is no heavy usage.

3

I would suspect the same. According to the thread dumps (thanks for those!) KNIME is busy doing zip operations. And I can imagine that some Windows security mechanism / virus scanner interferes with processes that do something with zip files.