Telcel · 2026
Moved the Telcel.com backend to a new cloud on Docker Swarm — then reinstalled it on RHEL 9.8 with zero regressions, proven by an evidence system I built.
- Client
- Telcel
- Role
- Solution architect · tech lead · sole operator
- Year
- 2026
Outcomes
Context
Telcel is Mexico's largest mobile carrier. The Telcel.com backend — back office, number portability, profiler, GeoServer and microsites, on legacy Java and Oracle — had to leave its old cloud. I ran the migration as sole operator: architecture, infrastructure, CI/CD, the operations handoff and QA. When the platform later needed a reinstall on RHEL 9.8, I built Datum to prove that nothing changed.
What I built
- 01
A Docker Swarm platform: 8 nodes, 12 stacks, 8 Jenkins pipelines, Nginx replacing HAProxy and Apache, 18 vhosts, 232 redirects and 16 microsites.
- 02
Observability and load testing: Grafana dashboards, alerting and k6 load tests.
- 03
Datum: a read-only system that captures an environment's baseline, compares the rebuild and classifies every difference with a citation and a decision. Runbooks, QA evidence and the client's status page come from the same data.
- 04
A live platform map for support: every machine, service and request path from domain to database — built only from GET-only reads of Docker, Prometheus, HTTP and TCP.
- 05
Before go-live it surfaced 31 missing alarms, 15 missing support accounts and a service running at half its replicas.