Troubleshooting Platform as a Service (PaaS) Issues
PaaS troubleshooting spans App Service, Azure Functions, Azure SQL, Cosmos DB, Storage, and messaging services. For App Service, separate plan-level problems (CPU/memory exhaustion, scale limits, quota) from app-level problems (bad configuration, missing app settings, startup failures, deployment slot swap effects). Know that Always On, WEBSITE_RUN_FROM_PACKAGE, and health check features change behavior; platform OS updates and regional outages appear in the service health blade.
Azure Functions traps: consumption-plan cold starts versus timeout limits; host.json and function.json binding errors surface as startup failures; missing storage account connectivity breaks triggers; and Durable Functions need a task hub storage account. SQL Database traps: firewall rules (server-level vs database-level), connection policy (proxy vs redirect), DTU/vCore throttling, and geo-replication failover requiring the app to update its connection string. Cosmos DB traps: partition key choice causing hot partitions, RU throttling (429), and consistency-level expectations. Storage traps: firewall and private endpoint restrictions, SAS expiry, and blob versioning or soft delete behavior.
Worked rule: when an app works locally but fails in Azure, check configuration first-app settings, connection strings, managed identity role assignments, and Key Vault references-then networking (private endpoints, access restrictions), then scaling limits. Read the platform diagnostic logs (App Service LogFiles, Kudu, Application Insights) before changing infrastructure.
Also distinguish transient throttling (retry with backoff) from persistent quota exhaustion (scale up/out or request quota).