A note on releases only happen at weekends
CI/CD for Endur: releasing the database and JVS together
Endur releases fail when the code and the database drift apart. A pipeline that versions and deploys them as one unit makes deployments boring again.
The thing you tested in UAT is byte-for-byte the thing that lands in production.
The single most common cause of a failed Endur deployment isn’t the JVS code and it isn’t the database — it’s the gap between them. Someone deploys the new scripts but misses a reference-data change, or runs the SQL but ships last week’s binaries. The fix is to stop treating them as separate things.
Version the database, not just the code
The first move is to bring the database under the same source control and the same versioning as everything else. Schema changes, reference-data seeds and configuration become ordered, idempotent migration scripts that live in the repo alongside the JVS. The pipeline never runs ad-hoc SQL against an environment; it only ever applies migrations in sequence, and it records which migration an environment is on.
This sounds obvious, and in most modern stacks it is. In ETRM it’s still surprisingly rare, because the database is often treated as a thing the DBA owns and the application as a thing the developers own. Collapsing that boundary is most of the battle.
One artefact, four components
A clean Endur release has a defined set of moving parts — database, interfaces, ATC and JVS among them. The pipeline should build all of them into a single, versioned, immutable artefact and promote that same artefact through dev, test and production. Octopus Deploy handles the promotion well: the build stage produces the package, and each environment is just a deployment of the identical package with environment-scoped variables.
The rule that makes this safe: the thing you tested in UAT is byte-for-byte the thing that lands in production. No rebuilds between environments, ever.
Make rollback a first-class step
Forward migrations are easy to write and easy to forget to reverse. Every database migration should ship with a tested down-path, and the deployment process should treat rollback as a normal, rehearsed operation rather than an emergency. The day you need it, you do not want to be writing the rollback script under pressure.
The payoff
Done well, an Endur release stops being a weekend event. It becomes a button: the pipeline applies the migrations, deploys the matching binaries, runs a smoke test, and reports green. The deployment is boring — which, for a trading platform, is the highest compliment you can pay it.
Is this happening on your desk?This is the consulting work I do. Tell me what you're seeing and I'll say whether I can help.
Send a consulting enquiry