Starting Session takes really slow - notebook session Microsoft Fabric

naanii - One thing I've noticed is that when using the default Workspace settings for Notebooks it always starts up lightning fast.  When I created a new default "Environment" to preinstall some libraries then it took longer to start-up the cluster.  < 30s  vs 2-3m.  Not sure if that is relevant to your situation or not.

We have this as well.  Default envirnoment is very quick, but custom environments are very slow.  It's quicker to %pip install all your stuff in the notebook rather than have it available to import from the envirnoment, which seems pretty mad to me.

Example:

Notebook with %pip install geopandas shapely pyproj fiona contextily, 11 seconds to start session and 28 seconds to run this first big import cell (installing and then importing all my libraries).

Notebook with in envirnoment with geopandas, shapely, pyproj, fiona and contextily installed there, and then import cell having the %pip part removed: 2 minutes to start session and then a further 20 seconds ish to run the cell. 

Hello,

I removed all the resources/items from My Workspace.

After creating a new notebook, I was able to start a session in about 5 sec.

Then I added a new Environment without any Libraries and this results in 3 minutes of starting time.

Why is this happing? We would like to import Libraries, but I think it is faster to import those in each of our notebooks instead of import those into a Environment.

Any ideas?

Was any solution/advice for this issue offered?

I'm seeing the exact same thing. A simple "Hello World" notebook using the default workspace environment will start a spark session in approx 4-5 seconds. Creating a new environment to add a custom library and setting it as the workspace default results in the same "Hello World" notebook now taking just over a minute to start a spark session.

This doesn't appear to be "Network latency, Regional Differences, browser issues, or temporary Fabric service hiccups..." as the difference in startup time can be reproduced with 100% reliability simply by switching the environment back and forth and re-running the same test notebook.

Spoke to one of the fantastic support people - sounds like it's unavoidable at the moment. Default environments are "ready to go", as soon as you create your own it takes time to spin them up.

Someone else has already submitted a request to improve this... hop on over and upvote if it'll help you: Microsoft Idea

I have also been facing a similair issue recently, with slow spark cluster session starts, even though we are still running on F64 trial.

Hi naanii , Anonymous ,

Thanks for using Fabric Community.
When I tried to check this issue, my session started in 5 sec.

Are you people still facing this issue?
Few factors can be - Network latency, Regional Differences, browser issues, or temporary Fabric service hiccups could also contribute to the delay.

Hope this is helpful. Please do let me know incase of further queries.

Hi naanii ,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet.
Otherwise, will respond back with the more details and we will try to help .

Hi all,

Since a couple of days it takes about 3min to start a session, where it previously was about 3sec!

I already switched to runtime 1.3 but the issue remains.

Location northern europe, trial version

Same issue here. I am opening a new notebook using default environment, the lakehouse I  added to the notebook is quite small, only a few thousands of sample data with 6 columns. But it took 2 minutes to create a connection and even to print out a small table takes 14 seconds.

I read from newspaper that Microsoft has taken measures to control cost related to cloud computation. I'm wondering if that explains the slow speed I'm experience. Does anyone experience the same issue recently?