Application Load Balancer 502, 503 and 504 errors
502, 503 and 504 from an ALB each point to a different problem between the load balancer and the targets. Match the code, then check health checks and timeouts.
What you see
HTTP 502: Bad gatewayHTTP 503: Service unavailableHTTP 504: Gateway timeout
What it means
These errors are generated by the load balancer, not your app. They are counted in the HTTPCode_ELB_5XX_Count metric. Errors your targets return themselves are counted separately in HTTPCode_Target_5XX_Count. That split tells you which side to look at first.
How to confirm the cause
- 1See target health and the failure reason
aws elbv2 describe-target-health --target-group-arn TARGET_GROUP_ARN
The console resource map labels failures as
Unhealthy: HTTP Response Mismatch,Unhealthy: Request timed outandUnhealthy: FailedHealthChecks. The docs list the full set under health check reason codes. - 2Check the load balancer access logs
# Enable access logs on the load balancer and read the error reason code on failing requests.
The docs point to the access log error reason codes for deeper detail.
Causes and fixes
502: target closed the connection or sent a bad response
- HOW TO TELL
- The target sent a TCP RST or FIN with a request outstanding, or the response was malformed.
- FIX
- Make sure the target's keep-alive timeout is longer than the load balancer's idle timeout, and that traffic is allowed from the load balancer subnets to the target port. Also check the target is not returning invalid headers.
502: Lambda target problem
- HOW TO TELL
- The target is a Lambda function that errored, timed out or returned a body over 1 MB.
- FIX
- Fix the function error, raise its timeout, or shrink the response.
503: no usable targets
- HOW TO TELL
- The target groups have no registered targets, or all are
unused. - FIX
- Register targets and make sure they are in the target group the listener rule routes to.
504: could not connect or the target was too slow
- HOW TO TELL
- Connection could not be established within 10 seconds, or the target did not answer before the idle timeout.
- FIX
- Check security groups and network ACLs allow traffic between the load balancer and targets, including the ephemeral ports 1024-65535. If the app is slow, raise the idle timeout or make the endpoint faster.
Targets are unhealthy
- HOW TO TELL
- The target shows
Unhealthy: HTTP Response Mismatch,Unhealthy: Request timed outorUnhealthy: FailedHealthChecks. - FIX
- For a response mismatch, make the app return the expected success code (200 by default) on the health check path, or change the expected codes. For a timeout, check security groups, ACLs, target capacity and use a simpler health check page.
How to stop it coming back
- Create a cheap dedicated health check endpoint.
- Set the app's keep-alive timeout above the load balancer idle timeout.
- Alert on
HTTPCode_ELB_5XX_Countseparately from target errors.
Keep going
- Runbook: ECS task stopped: how to find the reason
- Cheat sheet: AWS CLI
- Lesson: the networking topic in Learn
Sources
AWS documentation (docs.aws.amazon.com), read 18 Nov 2026.