I’m trying to install the v14 pipelines on a Jupyterhub server here in Edinburgh, but I’m getting an HTTP 403 error on the following URL: http://conda.lsst.codes/stack/0.14.0/linux-64/
I also tried the install using conda (which is more ideal for my setup as I’d like the pipelines to end up in the same Python environment as the rest of the Jupyter/Jupyterhub setup), but the instructions are gone: https://pipelines.lsst.io/install/conda.html and it also results in the same 403 error as the newinstall.sh method. Is conda still supported as an installation method? And if so, when will it be available for v14?
May I ask where that URL came from? It is not present in our 14.0 documentation and is not used by newinstall.
Conda packaging of LSST DM software was dropped after the 13.0 release as it wasn’t part of the developer workflow. This meant that changes in EUPS dependencies and new products would often result in the conda package build breaking. The conda package build was also attempting to replace EUPS packaged software with the official conda equivalent. This was also causing breakage as occasionally the official conda package was built (or rebuilt by upstream) differently than the EUPS equivalent or in some cases, was outright broken. At the time, the official conda packages were produced on centos 5 and some of our software needed patching to build on this platform, as it isn’t officially supported and not CI’d. The result was it took over a month of labor to produce the 13.0 conda package build. The conclusion was that conda packaging wasn’t sustainable unless it became the primary developer build mechanism. Our architecture team decided not to make that change.
We are producing EUPS tarball binary builds on a weekly (and unofficially, daily) basis. We also publish docker images from the EUPS tarballs into the docker.io registry weekly (and daily):
I’m not sure where that URL is coming from. I assumed it’s in the script, but it doesn’t appear to be there. Here’s my exact terminal output as I follow the instructions:
curl: (23) Failed writing body (757 != 2759)
!!! This script differs from the official version on the distribution
server. If this is not intentional, get the current version from here: https://raw.githubusercontent.com/lsst/lsst/master/scripts/newinstall.sh
Detected git version 2.11.0. OK.
In addition to Python 2 (>=2.7) or 3 (>=3.5), some LSST
packages depend on recent versions of numpy, matplotlib, and scipy. If you
do not have all of these, the installation may fail. Using the Miniconda
Python distribution will ensure all these are set up.
Would you like us to install the Miniconda Python distribution (if
unsure, say yes)?yes
######################################################################## 100.0%
Fetching package metadata …
The channel you requested is not available on the remote server.
You will need to adjust your conda configuration to proceed.
Use conda config --show to view your configuration’s current state.
Further configuration help can be found at http://conda.pydata.org/docs/config.html.
I’ve seen the Kubernetes/Jupyterlab demo stuff, thanks. It doesn’t apply to my case though as we’re not running Kubernetes or Open Stack. I’m just trying to get a working Jupyterhub installation on a bare metal machine that has the lsst stack loaded into the notebooks by default.
Can you please run the suggested conda config --show? I suspect that you have an outdated conda channel in your configuration that is being searched while installing normal conda packages. Removing that may fix things.