Skip to content

Latest commit

 

History

History
274 lines (195 loc) · 7.63 KB

README.md

File metadata and controls

274 lines (195 loc) · 7.63 KB
Raw

OpenShift deployment

At this time, the logdetective website is deployed in Fedora CommuniShift.

Tools

For some reason, in the Fedora repositories, there is an outdated (version 3.x) oc command (the origin-clients package). You might need to install it like this

https://docs.okd.io/latest/cli_reference/openshift_cli/getting-started-cli.html

Permissions

To be able to access the OpenShift project please ping fedora-infra to add you to communishift group and @FrostyX, to add you to communishift-logdetective group.

Login

First, log in using the OpenShift dashboard URL, then click your name in the top-right and "Copy login command". Display token and run the oc command.

oc login --token=... --server=https://api.fedora.cj14.p1.openshiftapps.com:6443

Build

The production container uses code from https://github.com/fedora-copr/logdetective-website so commit and push your changes.

Build the container image:

docker-compose -f docker-compose.prod.yaml build --no-cache

or

make build-prod

Push the image to quay.io quay.io/logdetective:

docker-compose -f docker-compose.prod.yaml push

or

make push-prod

Alternatively, when working with github repo, the image can be published by simply creating a tag with name respecting semver convention and starting with 'v' prefix, i.e. v0.0.4. This will trigger github action in docker-publish.yml.

New image will be created with the same tag and pushed to quay.io. Tag latest will also be updated to point to the new image.

The best way to publish an image is to make a new release of the project, using the github dialog. This will also generate release notes.

Update production

Once you created the new release, you should wait for our GitHub action to build a matching container image and push it to our quay repository.

When you see the image in the listing, continue with the regular Deploy process described below.

You can then verify based on the Image digest what we are running production:

oc describe pod logdetective-website-xxxxxxxx-xxxxx | grep 'Image ID'
   Image ID:       quay.io/logdetective/website@sha256:71a5cf95568e593df8a0723081b4bd17506d2236bdb431d9149f00337add3376

The Tag History shows these digests.

Pull Image

Alternativelly you can use existing image. Pre-build images are available at quay.io/logdetective, semantically versioned, with versions correponding to tags of this repo.

docker pull quay.io/logdetective/website:latest

Images are build and pushed with every new tag by a github action.

Deploy

Make sure you are using the correct OpenShift project

oc project communishift-logdetective

If a Kubernetes/OpenShift configuration change needs to be applied, run the following command. Otherwise you can skip it.

oc apply -f openshift/logdetective.yaml

To kill the current deployment and start a fresh, up-to-date container, run

oc rollout restart deploy/logdetective-website

You can debug the instance using

oc logs -f deploy/logdetective-website
# or
oc rsh deploy/logdetective-website

Secrets

Hugging Face token:

$ oc create secret generic hf-secret --from-literal=token=$TOKEN

TLS certificates

We use the certbot tool to get our Let's Encrypt certificates.

We generate a single certificate that works for both of our domains in 4 combinations:

  1. *.log-detective.com
  2. .log-detective.com
  3. *.logdetective.com
  4. .logdetective.com

We need all these 4 entries to get https://log... and https://www... working.

Both our domains are hosted on porkbun.com. We use DNS TXT entries to verify we own the domain during the cert generation process.

Ping @TomasTomecek to get access to the domain if you want to refresh the certs or set up a DNS entry.

Start by installing certbot:

$ dnf install certbot

Automated

There is a plugin certbot_dns_porkbun for certbot that makes domain refresh trivial.

Makes sure the plugin is installed in the same python sitelib as the main certbot tool. Otherwise certbot tool won't find it.

You can verify it works correctly like this:

$ certbot plugins

* dns-porkbun
Description: Obtain certificates using a DNS TXT record for Porkbun domains
Interfaces: Authenticator, Plugin
Entry point: EntryPoint(name='dns-porkbun',
value='certbot_dns_porkbun.cert.client:Authenticator', group='certbot.plugins')

The plugin uses porkbun's API secret and a key to authenticate. Once you have it, just fire this command and it should yield the certificates in a few minutes (DNS propagation can take some time, hence the 120 seconds below).

$ certbot certonly --config-dir cert/ --work-dir cert/ --logs-dir cert/ \
  --authenticator dns-porkbun --preferred-challenges dns --email $USER@redhat.com \
  -d '*.log-detective.com' -d '*.logdetective.com' \
  -d 'log-detective.com' -d 'logdetective.com' \
  --dns-porkbun-key $KEY \
  --dns-porkbun-secret $SECRET \
  --dns-porkbun-propagation-seconds 120

Head over now to the section Apply certificates.

Manual

This is a copy-pasta of Packit's process: https://github.com/packit/deployment/blob/main/docs/deployment/tls-certs.md Praise @jpopelka

Run certbot in the root of this git repo.

$ certbot certonly --config-dir cert/ --work-dir cert/ --logs-dir cert/ \
  --manual --preferred-challenges dns \
  --email $USER@redhat.com \
  -d '*.log-detective.com' -d '*.logdetective.com' \
  -d 'log-detective.com' -d 'logdetective.com'

You will soon be prompted:

Please deploy a DNS TXT record under the name:

Set those 2 TXT DNS entries for log-detective.com and logdetective.com and wait for DNS to resolve them:

$ watch -d nslookup -q=TXT _acme-challenge.logdetective.com

Alternatively check the record using porkbun's DNS server:

$ dig -t txt _acme-challenge.logdetective.com. @curitiba.ns.porkbun.com.

Delete those TXT DNS records.

All certificate stuff is in gitignored cert/ folder.

Apply certificates

You can verify the newly created cert with openssl CLI. Here we check that both domains are set as SAN:

$ openssl x509 -inform pem -noout -text -in 'cert/live/logdetective.com/fullchain.pem'
...
  X509v3 Subject Alternative Name:
      DNS:*.log-detective.com, DNS:*.logdetective.com, DNS:log-detective.com, DNS:logdetective.com

Make sure these paths point to the correct certificates:

  • "/persistent/letsencrypt/live/logdetective.com/cert.pem"
  • "/persistent/letsencrypt/live/log-detective.com/privkey.pem"
  • "/persistent/letsencrypt/live/log-detective.com/fullchain.pem"

The webserver expects them (see the production Dockerfile).

Copy the content of directory cert/ to the running logdetective website pod:

$ oc cp cert/ logdetective-website-$pod:/persistent

Connect to the pod, back up old certs and rename cert/ to letsencrypt/:

$ oc rsh deployment/logdetective-website

$ mv /persistent/letsencrypt{,-old}
$ mv /persistent/{cert,letsencrypt}

Kill the running pod so the certs are actually loaded.

🎉🎉🎉