Engineering decisions.
Explained in practice.
Practical guides to reliable systems, from tenant isolation and integrations to AI retrieval and product operations.
Page 5 of 7 · Articles 49–60 of 84
Checkout errors should offer a recovery path, not a dead end
A practical Product Design framework for classifying checkout failures, preserving user input, writing actionable errors, handling uncertain payments and measuring recovery.
OpenTelemetry tail sampling: keep evidence without overloading Collectors
A production guide to sizing and scaling OpenTelemetry tail sampling with trace-ID routing, decision windows, policy budgets, late-span handling and measurable overload protection.
Salla shipments: build a state machine that can recover
A production design for keeping Salla, a shipping carrier and an ERP consistent through duplicate webhooks, delayed events, unknown outcomes, cancellations and returns.
Product experiments need a decision contract
A practical product-management framework for defining hypotheses, success metrics, guardrails, data-quality checks and ship-or-stop rules before an experiment starts.
Kubernetes inference routing: balance tokens, not requests
A practical architecture for routing self-hosted LLM traffic with InferencePool and an Endpoint Picker using queue, cache and model signals instead of blind round robin.
Salla payment reconciliation: an order is not a ledger
A production design for reconciling Salla orders, payment transactions and refunds without treating a webhook or order status as complete accounting evidence.
PostgreSQL logical replication: monitor drift, not just lag
A production runbook for PostgreSQL logical replication covering apply conflicts, WAL retention, schema drift, reconciliation, slot failover and safe recovery.
Salla abandoned carts: build a recovery pipeline that knows when to stop
A production design for Salla abandoned-cart recovery using signed webhooks, current-state reads, durable suppression, reconciliation and external delivery providers.
Gemini 3.8 Live agents: design the session, not just the prompt
A production architecture for Gemini 3.8 Live covering WebSocket sessions, barge-in, async tools, reconnection, cost controls and durable business actions.
Zid bulk coupons: design the campaign before generating codes
A practical architecture for Zid bulk coupon campaigns covering rule snapshots, channel gates, secure distribution, cart validation and usage reconciliation.
Zid loyalty integrations: build around the ledger, not the balance
A practical architecture for Zid loyalty integrations using transaction history, command idempotency, read-back confirmation and scheduled reconciliation.
Kubernetes API Priority and Fairness: isolate noisy controllers
A practical APF design for protecting kube-apiserver from bursty controllers, operators and AI agents without starving critical traffic.