Conceptive

Replatforming

Replatforming a live portal without stopping the temple

A community runs bookings, donations and memberships every day. We rebuilt the platform underneath them in phases, and moved the data once.

18 September 2026, 4 min read

A major temple portal is not a typical e-commerce site. It manages high-volume darshan bookings, accommodation allotments, seva registrations, and financial donations. When quota releases happen, traffic spikes within seconds, and any digital failure immediately spills into the physical world, creating crowd-management problems at the gates. Because pilgrims book travel weeks in advance and walk-in devotees verify tickets continuously at ground counters, shutting the portal down for a prolonged maintenance window is out of the question. Replatforming must happen while every operational service remains live.

The cost of service interruption

Temple administrations face operational constraints that standard commercial systems rarely encounter:

  • Physical crowd safety: If ticket generation fails or validation services stall at entry turnstiles, queues back up into the streets. Latency in the software translates directly into crowd-control hazards on the ground.
  • Quota bursts: Darshan slots open at scheduled intervals. Tens of thousands of users attempt checkout simultaneously, requiring strict concurrency control and predictable database write performance.
  • Strict financial reconciliation: Donations, seva fees, and room tariffs must map precisely to trust accounting rules and bank settlement files without discrepancies.

A "big-bang" cutover carries unacceptable operational risk. If a new database locks under production concurrency or an integration point with on-premise scanners fails, the temple administration cannot simply roll back without losing records or stranding devotees. The replatforming strategy must decouple services incrementally.

Carving boundaries through the strangler pattern

Rather than rewriting the entire portal as a second monolith, engineering teams can place an API gateway or reverse proxy in front of the existing legacy portal. The gateway inspects incoming traffic and directs it based on domain boundaries.

The migration starts with read-heavy, low-risk services:

  • Static festival programmes and temple schedules
  • Information catalogues and dress-code rules
  • Devotional media and publications

Once the proxy routes these pages reliably to the new architecture, the engineering team tackles transactional boundaries. Isolating donor management and voluntary contributions provides a practical next step. Donations involve payment gateway interaction and immediate receipt generation, but they do not compete for finite inventory such as morning darshan slots. Running this module in production confirms that the new transaction processing, logging, and audit pipelines function correctly under real traffic.

Handling quota inventory and data synchronisation

The central challenge in a temple system is finite inventory. Accommodation rooms and special darshan slots cannot be overbooked. Allocating them across both legacy and modern platforms simultaneously introduces race conditions.

To maintain consistency during the migration, the inventory state must remain in a single source of truth:

  • Primary inventory ownership: Keep the core booking engine and slot allocation tables in the legacy database initially, even while modernising the public user interface.
  • Change data capture (CDC): Stream confirmed bookings from the primary database to the new platform using an event broker. The new platform uses this event stream to power reporting, real-time pilgrim notifications, and administrative dashboards.
  • Reconciliation over distributed commits: Avoid distributed transactions across old and new data stores. Instead, rely on idempotent consumers and automated reconciliation scripts that run every [figure needed] minutes to verify the booking ledger against payment gateway settlement files.

When the new booking engine proves it can handle the peak load of a scheduled quota release, the ownership of the inventory tables flips to the new database, turning the legacy system into a read-only store for historical audits.

Connecting the digital portal to physical entry points

A temple portal does not end in the browser; it extends to barcode scanners, handheld verification terminals, and turnstiles managed by ground staff.

During replatforming, these physical validation points must remain operational regardless of cloud network conditions:

  • Cryptographic ticket verification: Digital tickets should encode verifiable cryptographic signatures directly into the QR code. Handheld scanners at the gates can validate the signature offline or against a local cache, protecting the physical queue from network latency.
  • Dual validation support: On-premise gate software must be updated to recognise ticket identifiers generated by both the legacy portal and the new platform throughout the transition period.
  • Phased gate cutovers: Rather than shifting all physical entry lanes at once, ground validation begins with specific counters or a single gate. Ground operators verify hardware response times, barcode scanning speeds, and exception workflows before full rollout.

Replatforming a mission-critical public portal does not require shutting down access. By placing a routing layer ahead of legacy infrastructure, decoupling low-risk read paths, synchronising transactional state through event streams, and safeguarding ground-level validation, engineering teams modernise the stack predictably. The temple keeps its doors open, the queues continue moving, and the infrastructure updates without the public ever noticing the transition.

Have something to build, fix or connect?

Tell us the problem. We will tell you how we would approach it.