Automate deployments

I need the deployment for the payments service to run unattended tonight. Push the build artifact from the CI server to the production hosts, stop the current service, deploy the new release, run the DB migration script, start the service, and run the health check endpoints. Do it in that order on app1, app2, app3, with a rollback to the previous artifact if any step fails. Confirm completion by 23:00.

Before I hand this to the on-call engineer, make the deployment easy to validate. Put the version and expected downtime at the top, surface the single failing migration if present, make the service start logs and health-check output immediate to read, and flag any config drift that would block the release on app1, app2, app3.

We kicked off the automated deploy at 21:30 and app1 completed fine, but app2 aborted during the migration with a lock error. The team lead wants minimal downtime and is watching the customer SLA. I do not know if the lock is from a leftover process, a long-running transaction on the replica, or a migration script bug. What is the most likely cause and the fastest safe next move to get services back up without corrupting data?

We push deployments every evening and keep running into small failures on one host that take the morning to fix. We lose credibility with the product team and burn the on-call with late-night rollbacks. Where are we most likely wasting time or taking unnecessary risks in this pattern, and what single habit should we change to reduce these failures and restore predictable releases?