[aether-announce] New conda installation

[Andreas Hilboll]

Dear Aethers users,

you will have noticed that some weeks ago, we "improved" things on 
Aether which lead to your notebooks starting considerbly *slower*.

We have since identified the cause of the problem and as a 
consequence we are in the process of updating the conda/Python 
setup on Aether.


What's new about the LAMOS environments?
++++++++++++++++++++++++++++++++++++++++

The current environment lamos_2019.1 has been moved to a new 
location (`/home/conda/envs`).  The new JupyterHub installation is 
automatically using this environment, so that you won't have to do 
anything to start using it.  The package versions in this 
environment should not be affected by this move so all your code 
should continue running as usual.

Please update your `$HOME/.condarc` accordingly, i.e., make sure 
the the directory `/home/conda/envs` appears somewhere under 
`envs_dirs`.  If you need to use lamos_2018.1, you'll have to make 
sure to keep the directory `/opt/software/conda/envs` in there as 
well; otherwise you can (and should) remove it.  However, please 
be advised that we strongly encourage switching to lamos_2019.1 if 
your use case allows it.

We also created a new environment lamos_2019.2b (the `b` stands 
for "beta") with some additional packages (e.g., line_profiler, 
metpy, SkillMetrics, regionmask, holoviews/geoviews, glueviz, 
tensorflow, h5netcdf, geojson, ...).  Please feel free to try it 
out and report any issues you might have (see below on how to 
activate this environment).

The old lamos_2018.1 environment unfortunately cannot be moved due 
to technical reasons.  It will therefore remain in its current 
location, i.e., it will continue being rather slow in startup / 
importing.

The previous environment lamos_2018.1 will be deprecated (i.e., 
moved to the "optional" kernels, see below) once lamos_2019.2b 
will be stable (i.e., towards the beginning of the next semester).


How to manage your conda environments inside JupyterHub?
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

The current and previous LAMOS environments (as of writing, 
lamos_2019.1 and lamos_2018.1) will always be available in your 
JupyterHub.  You don't need to anything about this, we'll take 
care of it.

Additionally, we offer some more kernels which you can make 
available in your JupyterHub.  We don't make them visible by 
default in order to not clutter your JupyterHub launcher. 
Currently, the following additional kernels are available:

- lamos_2019.2b   -> This is the testing environment for the next 
  iteration of our "standard" environment.  It has updated 
  versions of many packages, plus some extras (see below).
- bash            -> A kernel for command line work.  Potentially 
  useful for communicating with your peers / supervisors about 
  what you're doing on the command line.
- ir_2016         -> A kernel for the R programming language which 
  we created back in 2016.
- python2-lpy0.2  -> A rather old Python2 kernel
- python2-lpy0.1  -> An even older Python2 kernel
- python3-lpy0.2  -> A rather old Python3 kernel
- python3-lpy0.1  -> An even older Python3 kernel

These kernels are defined in `/home/conda/kernels`.  When you log 
in to JupyterHub, it searches for kernels in 
`$HOME/.local/share/jupyter/kernels`.  This means that in order to 
make one of these optional kernels available in your JupyterHub, 
you simply need to create a symlink, e.g.,

    mkdir -p ~/.local/share/jupyter/kernels
    cd ~/.local/share/jupyter/kernels
    ln -s /home/conda/kernels/lamos_2019.2b

If at some point in the future, you don't want to have that kernel 
available any more, it is enough to remove that symlink,

    rm ~/.local/share/jupyter/kernels/lamos_2019.2b

Furthermore, if you have any local conda environments inside 
$HOME/.conda/envs which you want to be available in Aether 
JupyterHub, please follow the instructions at 
https://aether.uni-bremen.de/doc/python/#private-conda-environments.

In order for JupyterHub to "see" these changes, make sure to 
restart your JupyterHub server (not just one kernel).


We have also updated the documentation at 
https://aether.uni-bremen.de/doc/jupyter/ and 
https://aether.uni-bremen.de/doc/python/.  In case of any 
problems, please let uns know as soon as possible.


Greetings,
  Andreas

-- 
Dr. Andreas Hilboll

Institute of Environmental Physics (IUP)
University of Bremen

NW1 / S3132
Otto-Hahn-Allee 1
D-28359 Bremen
Germany

+49(0)421 218 62133  (phone)
+49(0)421 218 98 62133 (fax)
http://www.iup.uni-bremen.de/~hilboll