Problem
This was a production migration I led on a legacy PHP application serving roughly 100,000 daily active users, kept anonymized here.
Constraints
The goal was to move from PHP 7.3 to PHP 8.4 without turning a high-traffic system into a one-shot cutover. The application contained runtime assumptions and dependency paths that staging could not fully exercise.
What I did
Two Cloudflare Load Balancer pools ran side by side:
- the stable pool retained PHP 7.3 nodes
- the candidate pool ran PHP 8.4 nodes
The rollout began with 95% of traffic on PHP 7.3 and 5% on PHP 8.4. The candidate pool grew only as production behavior demonstrated that compatibility work was holding under real traffic.
Rollback
If a route failed on PHP 8.4 while remaining healthy on PHP 7.3, the rollback path was mechanical: disable the candidate pool, return traffic to the stable pool, isolate the compatibility issue, deploy the fix, and resume the canary.
Rollback was therefore a traffic-control operation, not an emergency deployment. The migration advanced in small, production-proven steps with a reversible path at every stage.
The important engineering decision wasn't installing the runtime — anyone can do that. It was designing a rollout where real traffic could surface unknown incompatibilities without turning them into a full production outage.
Result
The application moved from PHP 7.3 to PHP 8.4 while serving approximately 100,000 daily active users. The rollout began at 95% stable / 5% candidate, advancing as production behavior supported it.
Tradeoffs
Running two runtime pools adds operational work and requires careful comparison. In exchange, rollback is a traffic-control action and unknown incompatibilities can surface under limited exposure.