IonosatMicro PROMIS system deployment scripts and utilities

Navigate to the source directory and type vagrant up. It just works ™.

echo "development_setup: on" > conf/conf.yml
vagrant up

./sql_access

Runs a Postgres container that connects to the database and drops into the Postgres prompt. Can accept pipes and redirections to take SQL input. Must be run on a docker host obviously.

./backend_shell

Runs the Django shell in the backend container.

./backend_bash

Runs BASH in the backend container.

./backend_command <command>

Runs the Django command <command> in the backend container.

./backend_test

Runs tests inside the backend container.

./integration_test

Runs API and UI tests on the host machine (inside test container in future).

./backend_command loaddata --format json - < test/data/test_set.json

This does the job.

Check that your user is in the docker group. Alternatively force operation in VM (see below).

Synced folders and links persist through vagrant down/vagrant up even if you changed the config. Stop the container and remove it by docker's own means or use vagrant destroy <container-name to make sure you have new settings in effect if you are debugging. This also applies to inhouse containers being rebuilt because the id changes in process.

If you set the build option instead of image for a container, it will be built by vagrant itself. However, vagrant shall not automatically rebuild after the first build. To cause it to do so do vagrant reload.

Setting code_reload option (enabled automatically with development_setup) syncs major code directories of the git repository directly to the container allowing Django/Node.js to reload things on the fly. Note that you need to reload the container after changing this setting, because it's implemented as a shared folder.

On non-Linux machines vagrant would spin a boot2docker virtual machine. The default password is tcuser. Access the machine either through VirtualBox native means or do vagrant global-status, find the id of the running virtualbox machine corresponding to the docker host and do vagrant ssh <id>.

If a host VM is requested/required, folders are synced through virtualbox filesystem module. It is supposed to be slow, at least on OSX. In production environment this slowdown is not applicable due to folders being shared directly via docker.

If run under a VM, the port mapping in case ports specified are lower than 1000 is as follows:

• Web to 9080
• API (if different) to 9081

Set force_host_vm to true or on in conf/conf.yml.

There might be parallel build issues if you are using old vagrant. Try vagrant up --no-parallel, but really upgrade. 1.7 is quite buggy.

Copy conf/defaults.yml as conf/conf.yml and edit the values. You can only leave the options you want to override, the rest will be taken from defaults. You can reference conf.var_name in containers.yml.

Put your *.conf files in nginx/. They will not be picked up by Git because of .gitignore.

Everything that goes in ssl/ folder is mapped to /etc/ssl.d/ on the nginx container.

Override development_setup to on in conf.yml, it will turn SSL off, route hostnames to localhost and prefer local checkouts to git. The services will start as follows:

• https://localhost:8081 for Web
• https://localhost:8083 for API

Use this snippet for example. Run as ./genscr <hostname>, then <hostname>.csr will be your certificate ad <hostname>.key will be your private key. Put them to ssl/ and conf.yml.

Override expose_db to on in conf.yml, which gives you access to the Postgres at localhost:5432 (by default). This option is enabled automatically with development_setup.

Profiling is enabled automatically when django_debug is set in the config. The resulting profiles will be collected in the sync/profiling folder stating the path to the query and the total time it took. The files are saved in cProfile/pstats format and can be parsed as:

$ python3 -m pstat profile_file.prof
% help
% stats

The .prof files can be also visualized by SnakeViz. Just make sure it's Python3. Alternatively it's possible to convert the profiles to format acceptable for KCacheGrind/QCacheGrind:

$ pip3 install pyprof2calltree
$ for i in sync/profiling/*.prof; do pyprof2calltree -i $i -o callgrind.$i.txt; done

Don't enable --kcachegrind in runpromis.sh, it has an older version of the converter and seems to have a slight unit mixup.