README.md

SEDModS -- VLKB SED Models web service

This service is meant to provide filtering on top of the SED Models (big, 30M rows) catalogue table. The table was initially served as a RDBMS table and queried through SQL directly.

This work moves it onto a python+HDF5 solution.

The service is deployed as a python+WSGI on httpd, within a podman container to help maintenance.

Deploy SEDModS

Pull repo

When pulling this repo you'll end up with

|-- Dockerfile (container description)
|-- README.md (this text)
|-- sed-data
|   `-- link.to.hdf5 (dummy data file, see later)
|-- wsgi.conf (WSGI configuration for the http server)
`-- wsgi-scripts (actual service business logic)
    |-- hdf_query.py
    |-- query-server_d.py
    |-- wsgid.py
    `-- wsgi.py

Project content description

The Dockerfile cotains instructions to build the container that includes:

• an httpd server (running on container's port 80)
• python and WSGI packages to be able to run the service It needs a properly configured wsgi.conf that will be loaded within the http/conf.d of the container. The base container is a fedora:latest environment.

Besides the instructions to build the container and the WSGI configuration, the container, to properly run, needs two further pieces, two folders:

• the sed-data one that contains the SED Models HDF5 file
• the wsgi-scripts that actually contains the business logic of the service itself

(The actual HDF5 file is not kept in this repo, because it's too large and because it's not a wise choice to store byte-blobs in a versioning system).

The wsgi-scripts is provided and maintained in this repo, the sed-data one can be anywhere else, in reach of the host system. If the location of these these pieces has to be changed, it suffices to update accordingly the run command of the container (see below).

Build the container image

To build the container image, simply run

podman build --tag vlkb-sedmods -f ./Dockerfile

The tag can be one of your choice. It is suggested to have a dedicated user to run this in production.

Run the container

Once the podman image is ready and the two directories are in place, to run the container it suffices a command like

podman run -dt \
    --name sedmod-test \
    --rm \
    -v $PWD/sed-data:/sed-data:z \
    -v $PWD/wsgi-scripts:/var/www/wsgi-scripts:z \
    -p 8080:80 vlkb-sedmods

where the name is optional and of your choice and the left-hand side of the : in the -v arguments must point to the actual location of the two folders to be mounted in the container.

In the example the -v arguments are considered such as the command is run from the local working copy and the HDF5 SED Models file is actually within the sed-data folder.

Also, -p maps the 80 port of the container onto the 8080 port on the host server, this must be changed if the host's 8080 is already in use.

Service(s) in the container

Two flavours of the SEDModS service can run with the repo content:

• one that reads the HDF5 at each query
• one that reads the HDF5 once and works as a system daemon

Single query service mode

This mode is directly available when the cotainer is run. It uses the following code files:

wsgi.py
hdf_query.py

If you run on the host server

curl localhost:8080/sedsearch/?'clump_mass<10.005' > output.dat

you should get the response in the output.dat file, or you can point the browser to

http://host.server:8080/sedsearch/?'clump_mass<10.005'

and see the response directly.

Daemon service mode

This mode uses the code in:

wsgid.py
query_server_d.py

It requires a couple of processes to run before the deamonised service can work. These processes run within the container, so, after running it, one can launch them attaching to the running container with

podman exec -it sedmod-test /bin/bash

and within it run

python -m Pyro4.naming &
	python query-server_d.py &

After that, on can exit the shell and the daemon-based service should be reachable at

http://host.server:8080/seddaemon

with the same usage of the single query one.

Network Proxy

Since it could be annoying to explitly use port 8080 on the host, the service can be made visible on a specific context path in the host server's http using the httpd ProxyPass directive, like

<Location "/sedmods">
    ProxyPass "http://localhost:8080"
</Location>

where /sedmods is an example and the 8080 port depends on the passed parameters to the podman run command (see above).

SED Models HDF5 file

This is preserved, currently, on the INAF ICT Owncloud instance.

Dependencies

On the host:

• podman
• httpd

Within the container (i.e. provided in the build):

• httpd with python3-mod_wsgi
• python 3.x
• pandas
• Pyro4 (deamon mode)
• (py)tables