Jose Lausuch: Where you thinking about PCW while laying on the beach? 😛 Anton Smorodskyi: YES, constantly! :partygeeko: I see it in every palm on the beach !

PublicCloud-Watcher (PCW) is a web app which monitors, displays and deletes resources on various Cloud Service Providers (CSPs). PCW has three main flows :

1. Update run ( implemented in ocw/lib/db.py ) Executed every 45 minutes. Concentrates on deleting VMs (in case of Azure Resource Groups).

   • Each update scans accounts defined in configuration file and writes the obtained results into a local sqlite database. Newly discovered entities get assigned an obligatory time-to-life value (TTL). TTL may be taken from tag openqa_ttl if entity is tagged with such tag if not PCW will check pcw.ini for updaterun/default_ttl setting and if setting is not defined than PCW will use hard-coded value from webui/settings.py. Database has a web UI where you can manually trigger certain entity deletion.
   • After persisting results into db PCW deciding which entities needs to be deleted. There are two ways to survive for entity: a. Having tag pcw_ignore ( with any value) b. Age of entity is lower than TTL defined. Age is calculated as delta of last_seen and first_seen
   • For entities that survive cleanup PCW will sent notification email to the list defined in config.

2. Cleanup ( implemented in ocw/lib/cleanup.py ) Execution via django command. Concentrates on everything except VM deletion. This vary a lot per CSP so let's clarify that on per provider level.

   • For Azure such entities monitored (check details in ocw/lib/azure.py): a. bootdiagnostics b. Blobs in all containers c. Disks assigned to certain resource group defined in pcw.ini ('azure-storage-resourcegroup') d. Images assigned to certain resource group defined in pcw.ini ('azure-storage-resourcegroup') e. Image versions assigned to certain resource group defined in pcw.ini ('azure-storage-resourcegroup')
   • For EC2 such entities monitored (check details in ocw/lib/ec2.py): a. Images in all regions defined b. Snapshots in all region defined c. Volumes in all regions defined d. VPC's ( deletion of VPC means deletion of all assigned to VPC entities first ( security groups , networks etc. ))
   • For GCE deleting disks, images & network resources (check details in ocw/lib/gce.py)
   • For Openstack deleting instances, images & keypairs (check details in ocw/lib/openstack.py

3. Dump entities quantity ( implemented in ocw/lib/dumpstate.py ). To be able to react fast on possible bugs in PCW and/or unexpected creation of many resources there is ability to dump real time data from each CSP into defined InfluxDB instance. This allow building real-time dashboards and/or setup notification flow.

The fastest way to run PCW is via the provided containers, as described in the Running a container section.

PCW has 3 sets of virtual env requirements files :

• requirements.txt common usage for everything except K8S related cleanups
• requirements_k8s.txt due to high volume of dependencies needed only in single use case (k8s cleanups) they excluded in independent category
• requirements_test.txt contains dependencies allowing to run pcw's unit tests

Configuration of PCW happens via a global config file in /etc/pcw.ini. See templates/pcw.ini for a configuration template. To start, copy the template over:

    cp templates/pwc.ini /etc/pcw.ini

To be able to connect to CSP PCW needs Service Principal details. Depending on namespaces defined in pcw.ini PCW will expect some JSON files to be created under /var/pcw/[namespace name]/[Azure/EC2/GCE/Openstack].json. See templates/var/example_namespace/ for examples.

PCW supports email notifications about left-over instances. See the notify section therein and their corresponding comments.

# Setup virtual environment
virtualenv env
source env/bin/activate
pip install -r requirements.txt


## Configuration steps, only required once to setup the database and user
# Setup database
python manage.py migrate
# Setup superuser (OPTIONAL)
python manage.py createsuperuser --email [email protected] --username admin
python manage.py collectstatic


## Running the webapp server
python manage.py runserver

By default, PCW runs on http://127.0.0.1:8000/

In containers folder you main find several Dockerfiles to build several different images:

• Dockerfile image based on bci-python3.11 and can be used to run all PCW functionality except k8s cleanup
• Dockerfile_k8s image based on bci-python3.11 and can be used to run k8s cleanup
• Dockerfile_k8s_dev and Dockerfile_dev images which contains same set of dependencies as Dockerfile and Dockerfile_k8s and expect PCW source code to be mounted as volumes. Very usefull for development experiments

You can use the already build containers within this repository:

podman pull ghcr.io/suse/pcw_main:latest
podman pull ghcr.io/suse/pcw_k8s:latest

The PCW container supports two volumes to be mounted:

• (required) /etc/pcw.ini - configuration ini file
• (optional) /pcw/db - volume where the database file is stored

To create a container using e.g. the data directory /srv/pcw for both volumes and expose port 8000, run the following:

podman create --hostname pcw --name pcw -v /srv/pcw/pcw.ini:/etc/pcw.ini -v /srv/pcw/db:/pcw/db -v <local creds storage>:/var/pcw -p 8000:8000/tcp ghcr.io/suse/pcw_main:latest
podman start pcw

The pcw container runs by default the /pcw/container-startup startup helper script. You can interact with it by running

podman exec pcw /pcw/container-startup help

podman run -ti --rm --hostname pcw --name pcw -v /srv/pcw/pcw.ini:/etc/pcw.ini -v <local creds storage>:/var/pcw -v /srv/pcw/db:/pcw/db -p 8000:8000/tcp ghcr.io/suse/pcw_main:latest /pcw/container-startup help

To create an user within the created container named pcw, run

podman exec pcw /pcw/container-startup createuser admin USE_A_STRONG_PASSWORD

There is devel version of container file. Main difference is that source files are not copied into image but expected to be mounted via volume. This ease development in environment close as much as possible to production run.

Expected use would be :

make container-devel
podman run  -v <local path to ini file>:/etc/pcw.ini -v <local creds storage>:/var/pcw -v <path to this folder>:/pcw  -t pcw-devel "python3 manage.py <any command available>"

virtualenv .
source bin/activate
pip install -r requirements_test.txt
make test

The tests contain a Selenium test for the webUI that uses Podman. Make sure that you have the latest geckodriver installed anywhere in your PATH and that the podman.socket is enabled: systemctl --user enable --now podman.socket

Set the SKIP_SELENIUM environment variable when running pytest or make test to skip the Selenium test.

To simplify problem investigation pcw has several django commands :

cleanup

updaterun

dumpstate

rmclusters

those allows triggering core functionality without web UI. It is highly recommended to use dry_run = True in pcw.ini in such cases.