Define Release Strategy for Production

[erkannt]

Definition of Done:

• how and when do we release to production

Might require spin off of new issue to set up automation for this.

Probably depends on https://github.com/elifesciences/issues/issues/5519

[erkannt]

After a discussion with @hdrury1 and @Cooryd here:

When and how to deploy

• we automatically deploy to staging and production (already in place)
• we ensure that existing connections don't get terminated (Ensure that we can do rolling updates/replicas #998)
• gate for production deploy (Implement dev -> staging -> prod #944):
  • image versions to deploy match staging
  • staging has passed acceptance tests (Create acceptance tests to run against staging #923)
• reviewer deploy doesn't run journal end2end tests, but journal deploy does (Adjust journal end2end tests to work with libero reviewer #1080)
• production deploy must pass smoketest before receiving traffic (parity with xpub, most likely k8s health sufficient) (Create smoke tests to run against prod upon deployment #1042)

Alerting/Monitoring

• team to be alerted on failed staging and/or production releases (Set up Alerting for Production Deployment #791)
• these alerts are not #incident-channel, those pertain to issues with completed deploys
• monitoring of service after deploy is separate discussion (Setup monitoring/alerting for deployed service #1081)

[erkannt]

Where else to best capture this/share with journal tech team?

[erkannt]

Recorded in https://github.com/libero/reviewer/tree/master/docs