https://www.orchest.io/ logo
Title
k

Kay Kayyy

10/01/2021, 7:48 PM
hi - congrats on your funding! i saw it and decided to try it out, however the standard installation is failing. running ec2 ubuntu and your required version of docker. after it starts via CLI i hit the webapp and the home
/projects
is broken stuck in loading loop. chrome console showing a bunch of 500 errors. for example
/async/projects?session_counts=true&job_counts=true 500 (INTERNAL SERVER ERROR)
GET <http://IP:1122/catch/api-proxy/api/sessions/> 500 (INTERNAL SERVER ERROR)
Unable to fetch sessions Response {type: 'basic', url: '<http://IP:1122/catch/api-proxy/api/sessions/>', redirected: false, status: 500, ok: false, …}
🙌 1
y

Yannick

10/01/2021, 8:18 PM
Thank you! My hunch is that the volume size you chose is too small. Because Orchest requires a number of Docker containers (depending on installation it requires a maximum of 15GB last time I checked) you need a decent volume size. I recommend going for 50GB. Afterwards make sure to run the installation process again. If that doesn’t help, then please let me know. Thanks for raising the issue 🙏
k

Kay Kayyy

10/01/2021, 9:27 PM
thanks, i created a new ec2 with more space and now it works. would be great to add this to pre-reqs in docs or display some type of error about space, particularly during install-time rather than run-time. FWIW i had 7gb free. what exactly is happening where 7gb is not enough? just calling an API or listing projects requires more than 7gb of disk space?
r

Rick Lamers

10/01/2021, 9:27 PM
Also, to verify whether it’s working you can run
./orchest version --ext
It’s mainly the Jupyter Stacks base image required by the Jupyter Enterprise Gateway that balloon the installation size
Those are basically prebuilt images with a lot of open source data science packages. This ships as the default container image in which your DS code runs.
We’re always looking to reduce the initial installation size, but are somewhat dependent on upstream image use by the OSS we integrate
k

Kay Kayyy

10/01/2021, 9:30 PM
Jupyter Notebook instead of JupyterLab?
r

Rick Lamers

10/01/2021, 9:33 PM
That doesn’t make much difference in terms of size. It’s about the Docker image that provides the runtime environment for Python and R code of the user. In Jupyter context that means the kernels. Which are actually identical between Jupyter Notebook and JupyterLab.
k

Kay Kayyy

10/01/2021, 10:15 PM
another idea is instead of baking in & integrating an entire IDE, just have some plain text editor (
vi
equivalent)
r

Rick Lamers

10/01/2021, 10:17 PM
The IDE is actually pretty lightweight. It’s mainly the container images that make it larger. We do want to make everything more modular in the future allowing easier customization regarding what’s installed by default. For example, we already do this for the Julia, R and Python GPU base container images. They’re not included by default
They’re pulled on use