Test it first
On the test environment, with the version already bumped. If the application has no test environment, that is a request to the Deployment Lead — not a reason to skip ahead.
SolidPro Engineering Support · Deployment Committee
One route from a finished code change to a production deployment — tested on test, planned a day ahead, approved in writing, and logged. Every application, all six VPS, one trail that outlives the mail thread.
Scroll — the page moves sideways
All applications · all six VPS · Owner: Vinoth Kumar B, Deployment Lead
$ version bump
v2.4.0 → v2.4.1
$ deploy quotation-portal → vps-03
$
The whole rule set
On the test environment, with the version already bumped. If the application has no test environment, that is a request to the Deployment Lead — not a reason to skip ahead.
Production deployments are scheduled and planned the day before. Never on the same day they are asked for.
An approval mail to the Delivery Lead asking to deploy to production, CC the Deployment Lead and the BUH. Production waits for the reply.
In the DeployBoard, at the time you deploy. Test deployments too. The ledger is the record; nobody's memory of a Thursday afternoon is.
Production needs the Delivery Lead’s approval mail. Plan it one day ahead, and never deploy on the same day it is asked for. Test deployments need no approval — they still get the mail and the ledger entry.
The route
The spine is the normal deployment. Above it, the only path that may skip the test environment. Below it, what happens when there is no test environment to use. Pick any step to see why it is there.
Swipe the diagram sideways to follow it
Pick any step. Three decisions, two exception lanes, and one gate — the Delivery Lead's approval mail.
Before every production deployment
Not a notice — a request. Nothing reaches production until the Delivery Lead’s approval comes back, and it is asked for the day before, not the day of. Six lines, short enough to write from a phone.
The mail
To: Delivery LeadCc: Deployment Lead, BUHSubject: [PROD DEPLOY] <application> <version> -> <VPS> Application : quotation-portalVPS : vps-03Version : v2.4.1 (old: v2.4.0)Tested on test env : Yes, 19-AugWhat changed : one line
Inside
The portal is not a dashboard bolted onto a spreadsheet. Each screen owns one part of the record, and they all point at the same rows.
Sample data shown. Real figures live behind sign-in — a page anyone can load should not publish the size of the estate.
Latest release
quotation-portal
v2.4.1 · production
Where the last release stands, what is waiting, and what changed today — the first screen after sign-in.
Every release logged against the environment it went to, with the version it replaced. This is the column the Monday check reads.
What we run and where it runs — hosts, domains, certificates and the renewal dates that catch people out.
SSL certificate renewal
The steps for the jobs nobody does often enough to remember — renewals, restores, health checks. Written once, followed by anyone.
Minute taker
M. Iyer takes the minutes
Agenda, attendance, decisions and actions — and a minute taker drawn at random, so it is never the same person twice.
The management view: committed annual spend, what is on auto-renew, and what falls due in the next ninety days.
Weekly check
The VPS, the mail and the ledger are read against each other. A version running that nobody announced or logged is the finding this check exists to catch — not just a mail that never made it into the ledger. Found once, discussed once, and then it is over. The check exists to keep the record straight, not to find someone to blame.
2 gaps. Filled in on the call, discussed once, closed.
That is the whole process
The register, the ledger and the runbooks are all behind sign-in. Access is restricted to committee members — ask the Deployment Lead if you need an account.