Release 2.45.0

Release Engineering

Expectations for Release Engineer

A regularly scheduled release takes place Tuesday-Friday every other week at the end of a sprint.
Do not sign up for release engineer if you have any planned time off during Tues-Fri during the release week.
Testing on the Dogfood cluster should be complete by end of day Wednesday.
The release should be mostly complete by end of day Thursday. (No one wants to be troubleshooting release things on a Friday at 5 pm)
In order to expedite the release process, once a package upgrade is verified to have broken either the upgrade or the package UI functionality, the SOP will be to immediately back out the change for that package. usine the Package Rejection Process
join the Il4 Mattermost Big Bang Release Engineering Channel. This group chat will allow the Maintainers ¹ to support you. This group also provides the Maintainers ¹ with visibility on where we are in the release process.
The RE should seek anchor’s approval prior to starting step 1. Maintainers tend to push through last minute MRs on Tuesday morning. We want to get any finished work into the release wherever possible.
If a shadow is assigned, hop in a Zoom breakout room and pair to complete the release process.

Shadowing

Team members should shadow a release engineer prior to becoming RE. Assigned REs should have a solid understanding of the release process prior to their turn and have shadowed enough to understand and be able to execute the release process.
Try to clarify any questions about the release process steps prior to becoming the actual release engineer.
Review the release process steps before it's your turn. As always you can reach out to the Maintainers ¹ with questions about the release process.
Sync regularly with the release engineer to understand any issues/blockers that they run into. Likely you'll run into similar issues or blockers.

Prerequisites

Important: Make sure that you have met all prerequisites before the day you're scheduled to start working on a release in case you run into any problems.

Install or update all the tools listed in this Big Bang Training document
Install or update R2-D2 according to the instructions in the R2-D2 README
Install or update gogru according to the instructions in the gogru README
Request access to and/or ensure you can connect to the Dogfood cluster according to the instructions in the Dogfood README
Open Default Application Credentials
Have your P1 username, password and MFA device

Release Process

Forward

DO NOT blindly copy and paste commands prepended with $. Carefully read and evaluate these commands and execute them if applicable.
This upgrade process is applicable only to minor Big Bang version upgrades of Big Bang major version 2. It is subject to change in future major versions. Patch version upgrades follow a different process.
Use the Gogru release automation tool along with R2D2 to complete the release. Currently Gogru only supports the 1. Release Prep and 2. Upgrade and Debug Cluster parts of the release process. Follow the README for instructions on how to clone and install Gogru.
Use the R2D2 release automation tool to automate certain parts of this release process. Follow the README for instructions on how to clone and install R2D2. Usage of R2D2 is denoted in the document by the _R2-D2_ prefix.
The release branch format is as follows: release-<major>.<minor>.x where <minor> is the minor release number. Example: release-2.7.x. The <minor> string is a placeholder for the minor release number in this document.
The release tag format for major version 2 of Big Bang is: 2.<minor>.0.

NOTE: As you work through the release make a list of pain points / unclear steps. Once the release is complete, provide this feedback to the maintainers via MM and/or an MR to update the documentation. This allows us to continuously improve this process for future release engineers.

1. Release Prep

NOTE: If you need the manual steps and/or R2D2 steps, They can be found Here

Create the Bigbang (product) Issue
./gogru createIssue
Verify the Dogfood cluster is up and running normally
./gogru dogfoodCheck
- If there are any issues with the Helm Releases or Pods in the output, investigate and fix them before continuing
Create the release branches for Dogfood and BigBang (product), and verify the hash of the release branch matches the master branch
./gogru createRB
Update BigBang (product) version references with the new release version
./gogru update-version-references

2. Upgrade and Debug Cluster

WARNING: Upgrade only, do not delete and redeploy.

NOTE: If you need the manual steps, They can be found Here

Check and Renew ElasticSearch (ECK) License
./gogru eck-license-check
Check and Renew Mattermost License
./gogru mm-license-check
Upgrade Flux
./gogru update-flux
Upgrade Dogfood
./gogru upgrade-dogfood
- After running the upgrade-dogfood command, an MR will be opened in your name. Have the Maintainers ¹ merge your branch into master. Once merged, flux should pick up the changes and begin reconcile the helm releases. This may take up to an hour depending on the size of the release.
- As an extra check, you can watch the Release helmreleases, gitrepositories, pods, and kustomizations to check when all HRs have properly reconciled. Gogru's dogfoodCheck only looks at the HR's and Pods for now.
  watch kubectl get gitrepositories,kustomizations,hr,po -A
  IF FLUX HAS NOT UPDATED AFTER 10 MINUTES
  Shell Instructions to force Flux to reconcile
```shell flux reconcile hr -n bigbang bigbang --with-source --force ``` - If flux is still not updating, delete the flux source controller:
kubectl delete pod -n flux-system -l app=source-controller
- If the helm release shows max retries exhausted, check a describe of the HR. If it shows "another operation (install/upgrade/rollback) is in progress", this is an issue caused by too many Flux reconciles happening at once and typically the Helm controller crashing. You will need to delete helm release secrets and reconcile in flux as follows. Note that ${HR_NAME} is the same HR you described which is in a bad state (typically a specific package and NOT bigbang itself).
# example w/ kiali $ HR_NAME=kiali $ kubectl get secrets -n bigbang | grep ${HR_NAME}
# example output: # some hr names are duplicated w/ a dash in the middle, some are not sh.helm.release.v1.kiali-kiali.v1 helm.sh/release.v1 1 18h sh.helm.release.v1.kiali-kiali.v2 helm.sh/release.v1 1 17h sh.helm.release.v1.kiali-kiali.v3 helm.sh/release.v1 1 17m
# Delete the latest one: $ kubectl delete secret -n bigbang sh.helm.release.v1.${HR_NAME}-${HR_NAME}.v3 # suspend/resume the hr $ flux suspend hr -n bigbang ${HR_NAME} $ flux resume hr -n bigbang ${HR_NAME}
- If you see errors about no space left when you kubectl describe a failed deployment. The logs have probably filled the filesystem of the node. Determine which node the deployment was scheduled on and ssh to it and delete the logs. You need to have sshuttle running in order to reach the IP of the nodes.
ssh -i ~/.ssh/dogfood.pem ec2-user@xx.x.x.xx sudo -i rm -rf /var/log/containers/* rm -rf /var/log/pods/*
Then the deployment should recover on its own
Verify cluster has updated to the new release and verify that there are no outstanding issues with Pods or Helm Releases
./gogru dogfoodCheck

NOTE: If Prometheus does not start and is logging an error relating to Vault permissions, follow these instructions to reissue the Vault token.

NOTE: If for some reason the upgrade fails catastrophically and the release needs to be rolled back, follow the rollback instructions. This should be used only as a last resort to re-test the upgrade process.

HANDLING PACKAGE UPGRADE FAILURES

Follow the Package Rejection Process to rollback the package upgrade

3. UI Tests

WARNING: if you encounter a package that is failing please follow the Package Rejection Process to rollback the package upgrade.

NOTE: If you need the manual steps, They can be found Here

Default Application Credentials

Logging

Login to kibana with SSO. NOTE: If you get "You do not have permission to access the requested page," follow the instructions under "Renew ECK Trial" in the 2. Upgrade and Debug Cluster section above. Don't forget to log in as admin and reconfigure role mapping after you run the kubectl and flux commands.
Verify that Kibana is actively indexing/logging.
- To do this, click on the drop-down menu on the upper-left corner, then Under "Analytics" click Discover. Click "Create data view." In the "Index pattern" field, enter jaeger*. Set the "Timestamp field" to I don't want to use the time filter, then click "Use without saving". Log data should populate.
- If it is not indexing (you see no index for jaeger), you may have renewed the ElasticSearch license and jaeger did not reconcile after the renewal. Force jaeger to pick up the new configuration by doing: flux reconcile hr -n bigbang jaeger --force --with-source

Kiali

Login to kiali with SSO
Validate graphs and traces are visible under applications/workloads
- Graphs at the overview
- Graphs for inbound metrics
- Graphs for outbound metrics
- Graphs for Traces
Validate no errors appear
- NOTE: Red notification bell would be visible if there are errors. Errors on individual application listings for labels, etc are expected and OK.

Twistlock

Log in here
- In the Targets drop down choose serviceMonitor/twistlock/twistlock-console/0
- Click the small "show more" button that appears. You should see a State of UP and no errors.
R2-D2: Twistlock Test

Fortify

R2-D2: Fortify Test

GitLab Sonarqube Gitlab Runner

NOTE: if gitlab has updated to v18 and pipelines are not running, need to follow this procedure to register the runner Runner Registration, right now this process works but v18 gets rid of that bypass capability and we will have to follow a new process like here.

Login to gitlab with SSO
Edit profile and change user avatar. If your avatar doesn't change right away, check again later. It can take several minutes or more.
Go to Security > Access Token and create a personal access token. Select all scopes. Record the token. You'll need it for the next two steps.
R2-D2: Gitlab Test

Sonarqube

NOTE: If the Sonarqube is getting a major version update, you will need to go to 'http://sonarqube.dogfood.bigbang.mil/setup' to upgrade before you will be able to login.

Login to sonarqube with SSO
Verify your Project says PASSED

Nexus

R2-D2: Nexus Test

Minio

R2-D2: Minio Test

Mattermost

R2-D2: Mattermost Test

ArgoCD

R2-D2: ArgoCD Test

Kyverno

R2-D2: Kyverno Test

Velero

R2-D2: Velero Test
NOTE: This test fails occasionally. You may need to run it multiple times to get it to pass.

Loki

R2-D2: Loki Test
- NOTE: the Loki / Operational dashboard is currently missing and we have a task here to add it back in

Tempo

R2-D2: Tempo Test
- NOTE: this test is obsolete at the moment as we do not have any data tethered to tempo at the moment

Keycloak

R2-D2: Keycloak Test

Neuvector

R2-D2: Neuvector Test

Grafana & ArgoCD & Anchore & Mattermost SSO

R2-D2: Native SSO Grafana, ArgoCD, Anchore and Mattermost Test NOTE: Grafana OPA Violations dashboard showing "No Data" is a known issue. This has been skipped since 2.18.0, current open issue awaiting fix is here.

Monitoring & Cluster Auditor & Tracing (Jaeger) & Alertmanager

R2-D2: authservice SSO Prometheus, Tracing, AlertManager, and Cluster Auditor Test
- NOTE: Grafana OPA Violations dashboard showing "No Data" is a known issue. This has been skipped since 2.18.0, current open issue awaiting fix is here.
- NOTE: if vault shows up as UNHEALTHY, that means it was rebuilt and we need to update the permissions on the vault/secrets/token file within the prometheus pod. (we have an issue with the permissions on the file.)
  - to fix this:
    - exec into the vault-agent container on the prometheus-monitoring-monitoring-kube-prometheus-0 pod and run the following
```
chmod 644 /vault/secrets/token
```

Anchore

log back in as the admin user - password is in anchore-anchore-enterprise secret (admin will have pull credentials set up for the registries):

kubectl get secret anchore-anchore-enterprise -n anchore -o json | jq -r '.data.ANCHORE_ADMIN_PASSWORD' | base64 -d; echo ' <- password'

Scan image in dogfood registry, registry.dogfood.bigbang.mil/GROUPNAMEHERE/PROJECTNAMEHERE/alpine:latest
- Go to Images Page, Analyze Tag
  - Registry: registry.dogfood.bigbang.mil
  - Repository: GROUPNAME/PROJECTNAME/alpine
  - Tag: latest
Scan image in nexus registry, containers.dogfood.bigbang.mil/alpine:<release-number> (use your release number, ex: 1-XX-0). If authentication fails that means that the Nexus credentials have changed. Retrieve the Nexus credentials using instructions from Nexus above. Update creds in Anchore by clicking on menu Images > Analyze Repository > link at bottom "clicking here" > click on registry name > then edit the credentials.
Validate scans complete and Anchore displays data (click the SHA value for each image)

4. Create Release

STOP: Before finalizing the release, confirm that the latest bb-docs-compiler pipeline passed. If not, identify, correct, and cherry-pick a fix into the release branch. This will prevent wasting time re-running pipelines if the docs compiler pipeline fails.

Prep

Re-run helm docs to ensure that on the latest version + latest package versions.

cd bigbang
git pull
# pull any last minute cherry picks, verify nothing has greatly changed
git checkout release-2.<minor>.x
docker run -v "$(pwd):/helm-docs" -u $(id -u) jnorwood/helm-docs:v1.5.0 -s file -t .gitlab/base_config.md.gotmpl --dry-run > ./docs/understanding-bigbang/configuration/base-config.md
# commit and push the changes (if any)

Build The Release Notes

Create Release Candidate Tag

Create release candidate tag based on release branch, ie. 2.<minor>.0-rc.0. Tagging will additionally create release artifacts and the pipeline runs north of 1 hour. You will need to request tag permissions from the Maintainers ¹.
- To do this via the UI (generally preferred): tags page -> new tag, name: 2.<minor>.0-rc.0, create from: release-2.<minor>.x, message: "release candidate", release notes: leave empty
- To do this via git CLI:
```
git tag -a 2.<minor>.0-rc.0 -m "release candidate"
# list the tags to make sure you made the correct one
git tag -n
git push origin 2.<minor>.0-rc.0
```
Passed pipeline for Release Candidate tag.
Passed pipeline in bb-docs-compiler for latest RC tag (Gets created/scheduled towards the end of the bigbang release pipeline).

Review all pipeline output looking for failures or warnings. Reach out to the maintainers for a quick review before proceeding. Maintainers ¹
Check the releases for release notes.

Create Release Candidate Tag

Create release tag based on release branch. ie. 2.<minor>.0.
- To do this via the UI (generally preferred): tags page -> new tag, name: 2.<minor>.0, create from: release-2.<minor>.x, message: release 2.<minor>.0, release notes: leave empty
- To do this via git CLI:
```
git tag -a 2.<minor>.0 -m "release 2.<minor>.0"
# list the tags to make sure you made the correct one
git tag -n
# push
git push origin 2.<minor>.0
```
Passed release pipeline.

Review all pipeline output looking for failures or warnings. Reach out to the maintainers for a quick review before proceeding. Maintainers ¹ [ ]Ensure that the RC tag release created prior to this was deleted as well.
Ensure Release Notes are present here
Create a reminder to ensure that release docs have compiled and appear on Big Bang Docs and they have been mirrored to the IL2 Big Bang repository the following day.
- Expand the drop down in the title bar showing latest and confirm the current release is present.
- Click around in the docs and make sure docs are properly formatted, pay special attention to lists and tables.
NOTE: BB-docs won't actually publish the latest release version until the nightly bb-docs job runs. You can check this on the following day. NOTE: You may require access be granted to the IL2 Big Bang repository to view the status. Contact the Maintainers ¹ for assistance.

RELEASE IT

Type up a release announcement for the Maintainers ¹ to post in the Big Bang Value Stream channel. Mention any breaking changes and some of the key updates from this release. Use same format as a prior announcement https://chat.il2.dso.mil/platform-one/channels/team---big-bang

Cherry-pick release commit(s) as needed with merge request back to master branch. We do not ever merge release branch back to master. Instead, make a separate branch from master and cherry-pick the release commit(s) into it. Use the resulting MR to close the release issue.

# Navigate to your local clone of the BB repo
cd path/to/my/clone/of/bigbang
# Get the most up to date master
git checkout master
git pull
# Check out new branch for cherrypicking
git checkout -b 2.<minor>-cherrypick

# Repeat the following for whichever commits are required from release
# Typically this is the initial release commit (bumped GitRepo, Chart.yaml, CHANGELOG, README) and a final commit that re-ran helm-docs (if needed)
git cherry-pick <commit sha for Nth commit>

git push -u origin 2.<minor>-cherrypick
# Create an MR using Gitlab, merging this branch back to master, reach out to maintainers to review

Close Big Bang Milestone in GitLab. (make sure that there are n+2 milestones still, the current one and 1 future milestone, if not contact the maintainers ¹)
Handoff the release to the Maintainers ¹, they will review then celebrate and announce the release in the public MM channel
Reach out to @marknelson to let him know that the release is done and to update the Gitlab Banner

@jfoster @chris.oconnell @michaelmartin @andrewshoell ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³

57 of 77 checklist items completed · Edited 1 month ago

Designs

Child items 0

No child items are currently assigned. Use child items to break down this issue into smaller parts.

Release 2.45.0

Release Engineering

Expectations for Release Engineer

Shadowing

Prerequisites

Release Process

Forward

1. Release Prep

2. Upgrade and Debug Cluster

3. UI Tests

Default Application Credentials

Logging

Kiali

Twistlock

Fortify

GitLab Sonarqube Gitlab Runner

Sonarqube

Nexus

Minio

Mattermost

ArgoCD

Kyverno

Velero

Loki

Tempo

Keycloak

Neuvector

Grafana & ArgoCD & Anchore & Mattermost SSO

Monitoring & Cluster Auditor & Tracing (Jaeger) & Alertmanager

Anchore

4. Create Release

Prep

Build The Release Notes

Create Release Candidate Tag

Create Release Candidate Tag

RELEASE IT

Activity