Unable to start service

[MarkJM84]

Describe the bug
We are trying to setup the Qlik-Py-Tools connector. For this we use miniconda and the setup instructions on this page.
We have a working demo environment on virtual machines (local and Azure). But after a succesfull (no error) installation on our client's server we are unable to start the service.

Environment:

• OS: Windows Server 2016
• Qlik Sense Enterprise 13.82 (June 2020)
• SSE Version v.8.1

Additional context
We created a test server and used the follow (working) installation on our virtual machines:

1. Install Miniconda (www.conda.io)
   1a. Install for all users
   1b. Install miniconda to the %PATH%
2. Open Anaconda Prompt and execute the following commando's in here.
3. Downgrade python to 3.6.8
   Command: conda install python=3.6.8
4. Update conda
   Command: conda update conda
5. Install C++ compiler (Python: mingw-w64) as describe in the PyStan documentation
   Command: conda install libpython m2w64-toolchain -c msys2
6. Build the script
   Command: Qlik-Py-Init.bat
7. Add PyTools to the Analytic Connections in Qlik Sense
8. Start the SSE Plugin
   Command: Qlik-Py-Start.bat

Screenshots

[nabeel-oz]

Hi @MarkJM84, I haven't tried the install with miniconda and mingw. The installation guide is based on vanilla Python 3.6.8 and the Microsoft VS Build Tools.

You could compare your installation with this requirements file (pip freeze > requirements.txt) and see if there are any obvious discrepancies.

[MarkJM84]

Hi @nabeel-oz , thank you for answering.

There were some differences in our installation compared to the requirements file. But that shouldn't be the issue as it is the same installation as on our working test-machine.

I also did a reinstall of all the packages to the version number as mentioned in the requirement file, but even then we get the same error as in the first screenshot.

Current requirements:
requirements.txt

Error with Docker
We also tried to docker installation on our clients server, but weirdly enough we also have an error with that, might be connected.

Failed docker installation:

Machine specs

[Klief8788]

Hi @nabeel-oz,
I am facing the same issues on Windows Server 2016. I use conda to downgrade python to version 3.6.8 to meet the Prerequisites. In your reply you mentioned about vanilla Python. What is the difference between vanilla python and miniconda? How do we use Vanilla Python to install Qlik-Py-Tools.

[nabeel-oz]

Hi @Klief8788, by vanilla Python I meant the standard release from python.org that is referenced in the installation guide: https://github.com/nabeel-oz/qlik-py-tools#installation

I have not tried an installation with miniconda, and it did not work with anaconda the last time I tried, so I recommend using the standard Python 3.8 release: https://www.python.org/downloads/release/python-368/

[Klief8788]

Hi @nabeel-oz,
Thank you for your reply. We succeed to install Qlik-Py-Tools with miniconda on our own Azure virtual machine (tested on Windows Server 2016 and 2019. However it did not worked on the virtual machine of our customer. The difference is that this virtual machine is hosted by a Hyper-V cluster instead of Azure. We might think that the hardware virtualization causes this issue.

As recommended we have removed miniconda and installed Python 3.6.8 directly (downloaded from https://www.python.org/downloads/release/python-368/). We are now facing a Pystan issue (see attached image). Please note that we have installed the Visual C++ compilers.

[nabeel-oz]

Hi @Klief8788 , did you get any errors when running Qlik-Py-Init? If you did, we should discuss this under a new issue.

Almost all the pystan issues raised for this project have been due to C++ compilation errors, so I would double check that the required components of VS Build Tools have been installed. The components can be checked through Control Panel > Programs and Features > Visual Studio Build Tools and then choosing the Change option.

[Klief8788]

Hi @nabeel-oz,
No errors occurs when running Qlik-Py-Init.

The following components are installed on the virtual machine:

To install the required VS Build Tools components we checked the following boxes during the install:

[nabeel-oz]

Hi @Klief8788 , your screenshot shows Visual Studio Community instead of Build Tools so I'm not sure if all the required components are being installed. You can try selecting the following additional components:
MSVC v141 - VS 2017 C++ ARM64 build tools (v.14.16)
MSVC v142 - VS 2019 C++ ARM64 build tools (Latest)

The options seem to have changed since I wrote the installation guide which was based on Visual Studio Build Tools 2017. Here's a direct link to that version:
https://www.visualstudio.com/thank-you-downloading-visual-studio/?sku=BuildTools&rel=15

[Klief8788]

Hi @nabeel-oz,
We installed the additional components as suggested. Unfortunately the same error occurs when reinstalling Qlik_Py_Tools

Secondly we tried to install the VS c++ compilers from the Visual Studio Build Tools 2017. This does not solves the issue as well.

We rebooted the server before reinstalling Qlik_Py_Tools.
The Windows eventlog keeps reporting that the issue occurs on the following location:
C:\Python\qlik-py-tools-8.1\qlik-py-tools-8.1\qlik-py-env\Lib\site-packages\tensorflow\python\pywrap_tensorflow_internal.pyd

[nabeel-oz]

For VS 2017, try selecting the Visual C++ build tools workload in addition to the individual components. The screenshot does not show this workload as selected.

If the error is with tensorflow you could also be facing a processor incompatibility issue. Might be worth checking that the system is compatible with tensorflow 1.14.0 for CPUs.

[Klief8788]

Hi @nabeel-oz ,

As you can see we have selected the Visual C++ build tools components.

After that an error occured when running Qlik-Py-Init.bat. It reports that the module numpy.core._multiarray_umath' is not found.

The Qlik-Py-Start.bat reports that cygprc cannot be imported.

In the meanwhile we compared the CPU hardware of this environment with a working environment, which is hosted by Azure.
We might think that the CPU hardware does not meet the hardware requirements of Tensorflow.

CPU compare:
https://ark.intel.com/content/www/us/en/ark/compare.html?productIds=120506,64584