How to Fix AWS Application Load Balancer 504 Gateway Timeout Error
Quick Fix Summary
TL;DRIncrease the ALB idle timeout and verify your target's response time is under this limit.
A 504 Gateway Timeout from an AWS Application Load Balancer (ALB) indicates the ALB did not receive a response from its registered target (e.g., EC2, Lambda, ECS) within its configured idle timeout period. This is a client-side timeout from the ALB's perspective, not a failure from the target.
Diagnosis & Causes
Recovery Steps
Step 1: Check ALB & Target Group Metrics in CloudWatch
Identify if the timeout is at the ALB or target level. High `TargetResponseTime` and `HTTPCode_ELB_5XX_Count` confirm the issue.
aws cloudwatch get-metric-statistics --namespace AWS/ApplicationELB --metric-name TargetResponseTime --statistics Average --period 300 --start-time $(date -d '1 hour ago' +%Y-%m-%dT%H:%M:%SZ) --end-time $(date +%Y-%m-%dT%H:%M:%SZ) --dimensions Name=LoadBalancer,Value=app/your-alb-name/1234567890abcdef
aws cloudwatch get-metric-statistics --namespace AWS/ApplicationELB --metric-name HTTPCode_ELB_5XX_Count --statistics Sum --period 300 --start-time $(date -d '1 hour ago' +%Y-%m-%dT%H:%M:%SZ) --end-time $(date +%Y-%m-%dT%H:%M:%SZ) --dimensions Name=LoadBalancer,Value=app/your-alb-name/1234567890abcdefStep 2: Increase the ALB Idle Timeout
The default idle timeout is 60 seconds. Increase it if your application legitimately needs longer to respond. The maximum is 4000 seconds.
aws elbv2 modify-load-balancer-attributes --load-balancer-arn arn:aws:elasticloadbalancing:region:account-id:loadbalancer/app/your-alb-name/1234567890abcdef --attributes Key=idle_timeout.timeout_seconds,Value=120Step 3: Adjust the Target Group Deregistration Delay
When a target is failing, the ALB waits for in-flight requests to complete during the deregistration delay. Setting `deregistration_delay.timeout_seconds` too low can cause 504s during scaling events.
aws elbv2 modify-target-group-attributes --target-group-arn arn:aws:elasticloadbalancing:region:account-id:targetgroup/your-tg-name/abcdef1234567890 --attributes Key=deregistration_delay.timeout_seconds,Value=60Step 4: Profile Your Application's Response Time
Use application profiling (e.g., X-Ray, custom logging) to find slow database queries, external API calls, or code paths causing the delay.
# Enable X-Ray on your ALB (Console/CLI)
# Sample: Adding a response time header in a Node.js app (Express)
app.use((req, res, next) => {
const startHrTime = process.hrtime();
res.on('finish', () => {
const elapsedHrTime = process.hrtime(startHrTime);
const elapsedTimeInMs = elapsedHrTime[0] * 1000 + elapsedHrTime[1] / 1e6;
console.log(`Request Path: ${req.path}, Method: ${req.method}, Response Time: ${elapsedTimeInMs}ms`);
});
next();
});Step 5: Verify Target Health & Capacity
Ensure your backend instances (EC2, ECS tasks) have sufficient CPU/Memory and are passing health checks. An unhealthy target receiving requests will timeout.
aws elbv2 describe-target-health --target-group-arn arn:aws:elasticloadbalancing:region:account-id:targetgroup/your-tg-name/abcdef1234567890
# Check EC2 Instance Metrics
aws cloudwatch get-metric-statistics --namespace AWS/EC2 --metric-name CPUUtilization --dimensions Name=InstanceId,Value=i-1234567890abcdef0 --start-time $(date -d '1 hour ago' +%Y-%m-%dT%H:%M:%SZ) --end-time $(date +%Y-%m-%dT%H:%M:%SZ) --period 300 --statistics AverageStep 6: For Lambda Targets: Align Timeouts
If using Lambda, the ALB timeout must be greater than the Lambda function timeout. Also, ensure Lambda returns its response before timing out.
# 1. Set Lambda timeout (e.g., 30 sec) less than ALB timeout (e.g., 31 sec).
# 2. Verify Lambda does not exceed configured memory, causing throttling.Architect's Pro Tip
"Monitor the `UnHealthyHostCount` metric. A rising count with 504s often points to a backend capacity issue, not just a slow request."
Frequently Asked Questions
What's the difference between an ALB 504 and a 502 error?
A 504 (Gateway Timeout) means the ALB gave up waiting for your target. A 502 (Bad Gateway) means the target responded, but with an invalid or malformed HTTP response that the ALB could not process.
Can I set the ALB idle timeout to 0?
No. A value of 0 disables the idle timeout, which is not recommended for production as it can lead to resource exhaustion. The valid range is 1-4000 seconds.
My application is fast, but I still get sporadic 504s. Why?
This is often due to garbage collection pauses (Java/Python), cold starts (Lambda, containers), or TCP connection exhaustion between the ALB and your targets. Enable ALB access logs to see the `target_processing_time` field for outliers.