Open sqllensman opened 3 years ago
I have been trying to get this to work using a windows 10 hyper-v machine, the dev quick create version.
I failed to get the notebook experience working. I followed these steps
Download ADS 1.25.1 user from Download and install Azure Data Studio - Azure Data Studio | Microsoft Docs
Download Python 3.9.1 from Download Python | Python.org
Download notebook zip 6.2 from Releases · jupyter/notebook (github.com)
Download node.js from Download | Node.js
Create dir Create constraints.txt
Add backcall==0.1.0 bleach==1.5 html5lib==0.9999999 parso==0.3.1 tornado==4.2 and then 6.1 pyzmq==17.1
To text file
Run
py -m pip download -d ./jup_env/jupyterinstall jupyter==1.0.0 -c constraints.txt
Copy directory and all files to VM
Install, install, install
I then cd to the Jupyter install directory and ran
py -m pip install --no-index --find-links=c:\jupyterinstall jupyter
but this repeatedly failed at various points either due to tornado 4.2, tornado 6.1 or argon-cffi
at this point I gave up.
I also tried extracting the notebook zip and running
Cd to extracted path
python setup.py install --single-version-externally-managed --root=/
But that failed also.
@chlafreniere , @yualan , @markingmyname, @lucyzhang929 - FYI.
We have been investing in SQL & PowerShell notebooks to create a robust troubleshooting Book that can be used by our large (~50 person) DBA team to use to support our production environment. Some servers are ONLY accessible from secure, internet-disconnected RDP "jump servers" and any troubleshooting tools must be installed in a completely offline manner on those servers.
Additionally, some of the team is outsourced contractors, who depend on accessing internet-disconnected jump servers as their primary domain-joined access. (Using internet-connected virtual desktops isn't an option due to internal/political reasons).
Getting ADS up and running by itself is an OK experience -- downloading the installer + a number of vsix
packages is doable. However, getting PowerShell Notebooks to work is a seeming herculean feat. Similar to Mr Sewell's experience, the bird's nest of dependencies to get this minimally working has been (at best) discouraging, and frankly threatens the success of the solution for our team.
I have essentially done one-by-one "download, copy, install, try to open a notebook, parse errors, identify missing pip package, repeat" .... all the while attempting to keep track of the list of packages & versions so that we can automate this install on other jump servers. My concern is that even if I get this working, the maintenance & updates of all these python packages (50? 75? 100?) will be an untenable to maintain long-term.
After the upgrade to Version 1.30.0 the undisclosed requirement to upgrade to Python 3.8 resulted in my offline (working) version being ruined and being left in a state were Notebooks were again not usable.
As AzureData Studio is still not addressing the problems with installing in Offline or secure environments I will document below the steps I was able to take to actually get a working version. This is based on advice from https://stackoverflow.com/questions/11091623/how-to-install-packages-offline
First, on a machine that is open to Internet
In my case I installed to C:\Python\Python 3.8
Open an elevated Cmd session and navigate to the scripts directory of the Python Installation eg C:\Python\Python 3.8\Scripts
Check that pip is installed by running pip --version . This should return version of pip installed eg pip 21.1.1 from C:\Python\Python 3.8\lib\site-packages\pip <python 3.8>
Run the command pip freeze > requirements.txt
This will create a list of required packages
This will download libs and their dependencies to directory wheelhouse
On the machine without Internet Access
to a local directory
This will install the required packages onto the local Python Installation
Select Installation type of Use existing Python Installation and browse to C:\Python\Python 3.8 (or wherever you installed)
Continue the wizard. This should complete as all the required files are installed
Both the Python 3 and PowerShell Kernel's should now work.
Is your feature request related to a problem? Please describe. The current experience of installing Azure Data Studio on systems that are either behind Proxy servers or not connected to Internet is extremely time-consuming and frustrating. When trying to install or use Azure Data Studio on machines that are not left wide open to Internet the following issues always occur
Describe the solution or feature you'd like A clearly documented method to download and install ALL of required components and not just the main executable. The design of Azure Data Studio is such that almost all functionality requires additional components and these should be clearly listed and available to install New releases should clearly document when they are dependant on upgrades to other components.
Describe alternatives you've considered The only alternative outside of spending hours trying to resolve the poor offline installation experience is to simply abandon use entirely. Azure Data Studio without extensions and Python installed is basically unusable.
Additional context The same program that cannot install any of its own required extensions can successfully install the latest version of Powershell. After having to manually install Ver 1.2.4 I was able to update Powershell to V7.1 .