CloudConsole

RequestLimitExceeded: Request limit exceeded (EC2 API)

Updated 16 November 2026 · How we write and check guides

Short answer

The EC2 API throttles each account per region using token buckets, and you've used up the tokens for one category of calls. Back off and retry with jitter, then cut the call volume: filter Describe calls instead of listing everything, batch IDs into one request, and react to EventBridge state-change events instead of polling. If legitimate load still exceeds the limit, ask AWS Support for an increase.

Fix it, in 5 steps: 1. Retry with exponential backoff and jitter; 2. Ask for less; 3. Stop polling; 4. Spread the load; 5.…
The diagnosis and fix steps from this page, in the order to try them; commands and details are below.

What the error means

An error occurred (RequestLimitExceeded) when calling the DescribeInstances operation (reached max retries: 4): Request limit exceeded.

EC2 uses its own error code rather than the generic ThrottlingException; in CloudTrail it appears as Client.RequestLimitExceeded. Throttling works like a bucket of tokens per account and region: each call takes a token, tokens refill at a fixed rate, and short bursts can use up the bucket. Separate buckets apply to different kinds of calls, so heavy Describe* traffic doesn't use up the tokens for RunInstances, and vice versa. Resource-heavy actions like launching and terminating instances also have their own resource-rate limits.

What usually causes it

SourceWhy it adds up
Unfiltered DescribeInstances, DescribeVolumes or DescribeSnapshots in loopsLarge responses, paginated over many calls, repeated every few seconds
Waiters and status polling in scripts or pipelinesEach waiter polls on a fixed interval; dozens in parallel add up quickly
Kubernetes node autoscalers and CNI pluginsLarge clusters make many EC2 calls for nodes, ENIs and IP addresses
Terraform or inventory tools over big estatesOne refresh describes every resource in state
Several teams' tools in one accountAll share the same account-level buckets

Fix it

  1. Retry with exponential backoff and jitter. In SDKs and the CLI, set the standard or adaptive retry mode and allow more attempts:
    export AWS_RETRY_MODE=adaptive
    export AWS_MAX_ATTEMPTS=8
  2. Ask for less. Filter on the server and pass IDs in one call rather than one call per instance:
    aws ec2 describe-instances \
      --instance-ids i-0123456789abcdef0 i-0fedcba9876543210 \
      --query 'Reservations[].Instances[].[InstanceId,State.Name]' --output text
    
    aws ec2 describe-instances \
      --filters Name=tag:Service,Values=checkout Name=instance-state-name,Values=running
    --query only trims output on your side; --filters and IDs reduce the work and the pages.
  3. Stop polling. Use EventBridge rules on EC2 instance state-change notifications, or Auto Scaling lifecycle hooks, instead of loops that call Describe every few seconds.
  4. Spread the load. Stagger scheduled jobs that all start on the hour, and reduce concurrency in tools that make many parallel calls.
  5. Request an increase through AWS Support if, after the changes above, your normal workload still exceeds the limit. Include the API actions and request rates you need.

Find the noisy caller

CloudTrail records throttled EC2 calls. With CloudTrail in CloudWatch Logs:

filter eventSource = "ec2.amazonaws.com" and errorCode = "Client.RequestLimitExceeded"
| stats count(*) as throttled by eventName, userIdentity.arn, userAgent
| sort throttled desc

The userAgent field usually names the tool (Terraform, an autoscaler, a specific SDK), which tells you whom to talk to.

Mistakes to avoid

Don't add tight retry loops of your own on top of the SDK's retries; that multiplies the traffic that caused the problem. And don't move workloads into a second region just to escape throttling unless they belong there anyway. Fixing the polling pattern is cheaper and permanent.

Questions

What is the EC2 API request limit?

EC2 throttles API calls per account and region using token buckets with separate limits for different categories of actions. AWS documents the bucket sizes and refill rates, which can change, so check the EC2 API throttling documentation for current numbers.

Is RequestLimitExceeded the same as an instance limit?

No. It limits how fast you call the API. Running out of instances or vCPUs gives VcpuLimitExceeded or InsufficientInstanceCapacity instead.

Can I see when EC2 throttles me?

Yes. Throttled calls appear in CloudTrail with errorCode Client.RequestLimitExceeded, which you can count by caller with Logs Insights.

CloudConsole EC2 instance screen on a phone, for following up on RequestLimitExceeded away from your desk
When this happens away from your desk

EC2 throttling often hits during a scale-out, when launches and autoscaler calls spike at the same time. CloudConsole, a phone app for AWS, is built to show alarms and Auto Scaling activity on your phone so you can see whether capacity is actually arriving.

Did it start after a deploy? What changed? on the app’s alarm screen lists the deploys and configuration changes from the hour before the alarm fired.

About the app → · Download on the App Store

Sources