configuration

README

Red Hat Observability Service

This project holds the configuration files for our internal Red Hat Observability Service based on Observatorium.

See our website for more information about RHOBS.

Requirements

• Go 1.17+

macOS

• findutils (for GNU xargs)
• gnu-sed

Both can be installed using Homebrew: brew install gnu-sed findutils. Afterwards, update the SED and XARGS variables in the Makefile to use gsed and gxargs or replace them in your environment.

Usage

This repository contains Jsonnet configuration that allows generating Kubernetes objects that compose RHOBS service and its observability.

RHOBS service

The jsonnet files for RHOBS service can be found in services directory. In order to compose RHOBS Service we import many Jsonnet libraries from different open source repositories including kube-thanos for Thanos components, Observatorium for Observatorium, Minio, Memcached, Gubernator, Dex components, thanos-receive-controller for Thanos receive controller component, parca for Parca component, observatorium api for API component, observatorium up for up component, rules-objstore for rules-objstore component.

Currently, RHOBS components are rendered as OpenShift Templates that allows parameters. This is how we deploy to multiple clusters, sharing the same configuration core, but having different details like resources or names.

This is why there might be a gap between vanilla Observatorium and RHOBS. We have plans to resolve this gap in the future.

Running make manifests generates all required files into resources/services directory.

Observability

Similarly, in order to have observability (alerts, recording rules, dashboards) for our service we import mixins from various projects and compose all together in observability directory.

Running make prometheusrules grafana generates all required files into resources/observability directory.

Updating Dependencies

Up-to-date list of jsonnet dependencies can be found in jsonnetfile.json. Fetching all deps is done through make vendor_jsonnet utility.

To update a dependency, normally the process would be:

make vendor_jsonnet # This installs dependencies like `jb` thanks to Bingo project.
JB=`ls $(go env GOPATH)/bin/jb-* -t | head -1`

# Updates `kube-thanos` to master and sets the new hash in `jsonnetfile.lock.json`.
$JB update https://github.com/thanos-io/kube-thanos/jsonnet/kube-thanos@main

# Update all dependancies to master and sets the new hashes in `jsonnetfile.lock.json`.
$JB update

Testing cluster

The purpose of RHOBS testing cluster is to experiment before changes are rolled out to staging and production environments. The objects in the cluster are managed by app-interface, however the testing cluster uses a different set of namespaces - observatorium{-logs,-metrics,-traces}-testing.

Changes can be applied to the cluster manually, however they will be overridden by app-interface during the next deployment cycle.

Refresh token

The refresh token can be obtained via token-refresher.

./token-refresher --url=https://observatorium.apps.rhobs-testing.qqzf.p1.openshiftapps.com  --oidc.client-id=observatorium-rhobs-testing  --oidc.client-secret=<token> --log.level=debug --oidc.issuer-url=https://sso.redhat.com/auth/realms/redhat-external --oidc.audience=observatorium-telemeter-testing --file /tmp/token
cat /tmp/token

App Interface

Our deployments our managed by our Red Hat AppSRE team.

Updating Dashboards

Staging: Once the PR containing the dashboard changes is merged to main it goes directly to stage environment - because the telemeter-dashboards resourceTemplate refers the main branch here.

Production: Update the commit hash ref in the saas file in the telemeterDashboards resourceTemplate, for production environment.

Prometheus Rules and Alerts

Use synchronize.sh to create a MR against app-interface to update dashboards.

Components - Deployments, ServiceMonitors, ConfigMaps etc...

Staging: update the commit hash ref in https://gitlab.cee.redhat.com/service/app-interface/blob/master/data/services/telemeter/cicd/saas.yaml

Production: update the commit hash ref in https://gitlab.cee.redhat.com/service/app-interface/blob/master/data/services/telemeter/cicd/saas.yaml

CI Jobs

Jobs runs are posted in:

#sd-app-sre-info for grafana dashboards

and

#team-monitoring-info for everything else.

Troubleshooting

1. Enable port forwarding for a user - example
2. Add a pod name to the allowed list for port forwarding - example