Production Deployment

Learn more about Mintlify

Enter your email to receive updates about new features and product releases.

Overview
Performance Optimization
Connection Pooling
Query Complexity Limits
Caching Strategies
HTTP Server Tuning
Monitoring and Observability
Metrics Collection
Error Reporting
Logging
Distributed Tracing
Error Handling
Custom Error Responses
Graceful Degradation
Security
Authentication
Authorization
Rate Limiting
Input Validation
Scaling
Horizontal Scaling
Load Balancing
Auto-scaling
Health Checks
Deployment Strategies
Blue-Green Deployment
Canary Deployment
Rolling Update
Configuration Management
Checklist
Next Steps

Overview

Deploying Viaduct to production requires careful consideration of performance, monitoring, error handling, and scalability. This guide covers best practices and production-specific concerns.

Performance Optimization

Connection Pooling

Configure connection pooling for all data sources to avoid connection overhead:

import com.zaxxer.hikari.HikariConfig
import com.zaxxer.hikari.HikariDataSource

val config = HikariConfig().apply {
    jdbcUrl = "jdbc:postgresql://localhost:5432/mydb"
    username = "user"
    password = "password"
    maximumPoolSize = 20
    minimumIdle = 5
    connectionTimeout = 30000
    idleTimeout = 600000
    maxLifetime = 1800000
}

val dataSource = HikariDataSource(config)

Query Complexity Limits

Protect your API from expensive queries:

import graphql.analysis.MaxQueryComplexityInstrumentation
import graphql.analysis.MaxQueryDepthInstrumentation

val viaduct = ViaductBuilder()
    .withMaxQueryComplexity(1000)  // Limit total complexity
    .withMaxQueryDepth(15)         // Limit nesting depth
    .build()

Caching Strategies

Implement caching at multiple levels: 1. DataLoader for batch loading:

import org.dataloader.DataLoader
import org.dataloader.DataLoaderRegistry

val userLoader = DataLoader.newDataLoader<String, User> { ids ->
    CompletableFuture.supplyAsync {
        userRepository.findByIds(ids)
    }
}

val registry = DataLoaderRegistry()
registry.register("users", userLoader)

2. Response caching:

// Cache entire query results for common queries
val queryCache = ConcurrentHashMap<String, ExecutionResult>()

val cacheKey = "${query.hashCode()}_${variables.hashCode()}"
val cachedResult = queryCache[cacheKey]
if (cachedResult != null && !isStale(cachedResult)) {
    return cachedResult
}

val result = viaduct.executeAsync(executionInput).await()
queryCache[cacheKey] = result

3. Field-level caching:

import com.github.benmanes.caffeine.cache.Caffeine
import java.util.concurrent.TimeUnit

val fieldCache = Caffeine.newBuilder()
    .expireAfterWrite(5, TimeUnit.MINUTES)
    .maximumSize(10000)
    .build<String, Any>()

HTTP Server Tuning

Configure your HTTP server for production load: Jetty:

import org.eclipse.jetty.server.Server
import org.eclipse.jetty.util.thread.QueuedThreadPool

val threadPool = QueuedThreadPool().apply {
    minThreads = 10
    maxThreads = 200
    idleTimeout = 60000
}

val server = Server(threadPool)

Ktor:

import io.ktor.server.engine.embeddedServer
import io.ktor.server.jetty.Jetty

embeddedServer(Jetty, port = 8080) {
    // ...
}.start(wait = true)

Monitoring and Observability

Metrics Collection

Integrate with Micrometer for comprehensive metrics:

import io.micrometer.core.instrument.MeterRegistry
import io.micrometer.prometheus.PrometheusConfig
import io.micrometer.prometheus.PrometheusMeterRegistry

val meterRegistry: MeterRegistry = PrometheusMeterRegistry(PrometheusConfig.DEFAULT)

val viaduct = ViaductBuilder()
    .withMeterRegistry(meterRegistry)
    .build()

Key metrics to monitor:

Query execution time (percentiles: p50, p95, p99)
Error rate by query type
Resolver execution time
Data source connection pool utilization
Query complexity distribution
Cache hit/miss ratio
Concurrent request count
JVM memory and GC metrics

Error Reporting

Implement comprehensive error reporting:

import viaduct.service.api.spi.ErrorReporter
import graphql.GraphQLError

class ProductionErrorReporter(
    private val sentryClient: SentryClient,
    private val logger: Logger
) : ErrorReporter {
    override fun report(error: Throwable, context: Map<String, Any>) {
        // Log error locally
        logger.error("Resolver error: ${error.message}", error)
        
        // Send to error tracking service
        sentryClient.sendException(error, mapOf(
            "query" to context["query"],
            "variables" to context["variables"],
            "userId" to context["userId"]
        ))
    }
    
    override fun reportGraphQLError(error: GraphQLError, context: Map<String, Any>) {
        logger.warn("GraphQL error: ${error.message} at ${error.path}")
    }
}

val viaduct = ViaductBuilder()
    .withResolverErrorReporter(ProductionErrorReporter(sentryClient, logger))
    .build()

Logging

Configure structured logging:

import org.slf4j.LoggerFactory
import net.logstash.logback.argument.StructuredArguments.*

val logger = LoggerFactory.getLogger("ViaductAPI")

logger.info(
    "GraphQL query executed",
    keyValue("operationName", operationName),
    keyValue("userId", userId),
    keyValue("duration", durationMs),
    keyValue("complexity", queryComplexity),
    keyValue("hasErrors", result.errors.isNotEmpty())
)

Log levels:

ERROR: Unhandled exceptions, system failures
WARN: GraphQL validation errors, deprecated field usage
INFO: Query execution metrics, authentication events
DEBUG: Detailed query execution, resolver invocations (avoid in production)

Distributed Tracing

Implement distributed tracing for microservices:

import io.opentelemetry.api.trace.Tracer
import io.opentelemetry.api.trace.Span

class TracingResolver(
    private val tracer: Tracer,
    private val userService: UserService
) {
    fun getUser(id: String): User? {
        val span = tracer.spanBuilder("resolver.user").startSpan()
        try {
            span.setAttribute("userId", id)
            return userService.findById(id)
        } finally {
            span.end()
        }
    }
}

Error Handling

Custom Error Responses

Provide user-friendly error messages while logging details:

import viaduct.service.api.spi.ResolverErrorBuilder
import graphql.GraphQLError
import graphql.language.SourceLocation

class ProductionErrorBuilder : ResolverErrorBuilder {
    override fun build(
        exception: Throwable,
        path: List<Any>,
        locations: List<SourceLocation>
    ): GraphQLError {
        // Log internal details
        logger.error("Resolver error at path: $path", exception)
        
        // Return sanitized error to client
        return when (exception) {
            is ValidationException -> 
                GraphQLError.newError()
                    .message(exception.message)
                    .path(path)
                    .locations(locations)
                    .build()
            
            is AuthenticationException ->
                GraphQLError.newError()
                    .message("Authentication required")
                    .extensions(mapOf("code" to "UNAUTHENTICATED"))
                    .build()
            
            is AuthorizationException ->
                GraphQLError.newError()
                    .message("Permission denied")
                    .extensions(mapOf("code" to "FORBIDDEN"))
                    .build()
            
            else ->
                GraphQLError.newError()
                    .message("An internal error occurred")
                    .extensions(mapOf(
                        "code" to "INTERNAL_ERROR",
                        "timestamp" to System.currentTimeMillis()
                    ))
                    .build()
        }
    }
}

val viaduct = ViaductBuilder()
    .withDataFetcherErrorBuilder(ProductionErrorBuilder())
    .build()

Graceful Degradation

Handle partial failures gracefully:

fun getUser(id: String): User? {
    return try {
        userService.findById(id)
    } catch (e: DatabaseException) {
        logger.error("Database error fetching user $id", e)
        // Return cached data if available
        userCache[id] ?: throw e
    }
}

Security

Authentication

Implement authentication at the HTTP layer:

import io.jsonwebtoken.Jwts

fun authenticateRequest(authHeader: String?): User? {
    val token = authHeader?.removePrefix("Bearer ") ?: return null
    
    return try {
        val claims = Jwts.parserBuilder()
            .setSigningKey(jwtSecret)
            .build()
            .parseClaimsJws(token)
            .body
        
        val userId = claims["userId"] as String
        userService.findById(userId)
    } catch (e: Exception) {
        logger.warn("Invalid authentication token", e)
        null
    }
}

// In your HTTP handler
val user = authenticateRequest(request.headers["Authorization"])
if (user == null) {
    return respondUnauthorized()
}

val executionInput = ExecutionInput.create(
    operationText = query,
    variables = variables,
    requestContext = mapOf("currentUser" to user)
)

Authorization

Implement field-level authorization in resolvers:

import graphql.schema.DataFetchingEnvironment

fun sensitiveField(env: DataFetchingEnvironment): String? {
    val currentUser = env.graphQlContext.get<User>("currentUser")
    
    if (!currentUser.hasPermission("read:sensitive_data")) {
        throw AuthorizationException("Insufficient permissions")
    }
    
    return fetchSensitiveData()
}

Rate Limiting

Protect against abuse:

import io.github.bucket4j.Bucket
import io.github.bucket4j.Bandwidth
import java.time.Duration

val userBuckets = ConcurrentHashMap<String, Bucket>()

fun rateLimitCheck(userId: String): Boolean {
    val bucket = userBuckets.computeIfAbsent(userId) {
        Bucket.builder()
            .addLimit(Bandwidth.simple(1000, Duration.ofHours(1)))
            .build()
    }
    
    return bucket.tryConsume(1)
}

// In your HTTP handler
if (!rateLimitCheck(currentUser.id)) {
    return respondTooManyRequests()
}

Input Validation

Validate all inputs:

fun createUser(input: CreateUserInput): User {
    require(input.email.matches(emailRegex)) {
        "Invalid email format"
    }
    
    require(input.name.length in 1..100) {
        "Name must be between 1 and 100 characters"
    }
    
    return userService.create(input)
}

Scaling

Horizontal Scaling

Viaduct instances are stateless and can be scaled horizontally:

# Kubernetes deployment example
apiVersion: apps/v1
kind: Deployment
metadata:
  name: viaduct-api
spec:
  replicas: 5  # Scale to 5 instances
  selector:
    matchLabels:
      app: viaduct-api
  template:
    metadata:
      labels:
        app: viaduct-api
    spec:
      containers:
      - name: viaduct-api
        image: myregistry/viaduct-api:latest
        ports:
        - containerPort: 8080
        resources:
          requests:
            memory: "512Mi"
            cpu: "500m"
          limits:
            memory: "1Gi"
            cpu: "1000m"
        livenessProbe:
          httpGet:
            path: /health
            port: 8080
          initialDelaySeconds: 30
          periodSeconds: 10
        readinessProbe:
          httpGet:
            path: /health
            port: 8080
          initialDelaySeconds: 5
          periodSeconds: 5

Load Balancing

Configure load balancing:

# Kubernetes service
apiVersion: v1
kind: Service
metadata:
  name: viaduct-api
spec:
  selector:
    app: viaduct-api
  ports:
  - protocol: TCP
    port: 80
    targetPort: 8080
  type: LoadBalancer

Auto-scaling

Configure horizontal pod autoscaling:

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: viaduct-api
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: viaduct-api
  minReplicas: 3
  maxReplicas: 20
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 70
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 80

Health Checks

Implement comprehensive health checks:

data class HealthStatus(
    val status: String,
    val version: String,
    val uptime: Long,
    val checks: Map<String, Boolean>
)

fun healthCheck(): HealthStatus {
    val checks = mutableMapOf<String, Boolean>()
    
    // Check database connectivity
    checks["database"] = try {
        dataSource.connection.use { it.isValid(5) }
    } catch (e: Exception) {
        false
    }
    
    // Check cache connectivity
    checks["cache"] = try {
        redisClient.ping()
        true
    } catch (e: Exception) {
        false
    }
    
    // Check external services
    checks["authService"] = authService.isHealthy()
    
    val isHealthy = checks.values.all { it }
    
    return HealthStatus(
        status = if (isHealthy) "UP" else "DOWN",
        version = BuildConfig.VERSION,
        uptime = System.currentTimeMillis() - startTime,
        checks = checks
    )
}

Deployment Strategies

Blue-Green Deployment

Deploy new version to green environment
Run smoke tests against green
Switch load balancer to green
Keep blue environment as rollback option
Shut down blue after stability confirmed

Canary Deployment

Deploy new version to small subset of instances (5-10%)
Monitor error rates and performance
Gradually increase traffic to new version
Roll back if issues detected
Complete rollout if metrics are healthy

Rolling Update

apiVersion: apps/v1
kind: Deployment
metadata:
  name: viaduct-api
spec:
  replicas: 10
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxSurge: 2        # Add 2 new pods before removing old
      maxUnavailable: 1  # Allow 1 pod to be down during update

Configuration Management

Use environment variables for configuration:

val config = object {
    val databaseUrl = System.getenv("DATABASE_URL") 
        ?: throw IllegalStateException("DATABASE_URL not set")
    val jwtSecret = System.getenv("JWT_SECRET")
        ?: throw IllegalStateException("JWT_SECRET not set")
    val redisUrl = System.getenv("REDIS_URL")
    val logLevel = System.getenv("LOG_LEVEL") ?: "INFO"
    val environment = System.getenv("ENVIRONMENT") ?: "production"
}

Checklist

Before going to production:

Next Steps

Embedding Viaduct

Review the embedding guide

Development Server

Learn about local development

Monitoring

Deep dive into monitoring and observability

Security

Learn more about security best practices

Development Server

⌘I

Build docs developers (and LLMs) love

Get started for free Talk to us

Get Started

Core Concepts

Building Your Tenant

Advanced Topics

Deployment

​Overview

​Performance Optimization

​Connection Pooling

​Query Complexity Limits

​Caching Strategies

​HTTP Server Tuning

​Monitoring and Observability

​Metrics Collection

​Error Reporting

​Logging

​Distributed Tracing

​Error Handling

​Custom Error Responses

​Graceful Degradation

​Security

​Authentication

​Authorization

​Rate Limiting

​Input Validation

​Scaling

​Horizontal Scaling

​Load Balancing

​Auto-scaling

​Health Checks

​Deployment Strategies

​Blue-Green Deployment

​Canary Deployment

​Rolling Update

​Configuration Management

​Checklist

​Next Steps

Embedding Viaduct

Development Server

Monitoring

Security

Build docs developers (and LLMs) love

Overview

Performance Optimization

Connection Pooling

Query Complexity Limits

Caching Strategies

HTTP Server Tuning

Monitoring and Observability

Metrics Collection

Error Reporting

Logging

Distributed Tracing

Error Handling

Custom Error Responses

Graceful Degradation

Security

Authentication

Authorization

Rate Limiting

Input Validation

Scaling

Horizontal Scaling

Load Balancing

Auto-scaling

Health Checks

Deployment Strategies

Blue-Green Deployment

Canary Deployment

Rolling Update

Configuration Management

Checklist

Next Steps