ResourceInitializationError: unable to pull secrets or registry auth
Before starting your container, ECS couldn't fetch something it needs on the task's behalf: secrets from Secrets Manager or Parameter Store, or the ECR authorisation token. The cause is almost always either no network path from the task's subnets to those services, or a task execution role missing secretsmanager:GetSecretValue, ssm:GetParameters or kms:Decrypt for the referenced secret.
What the message looks like
ResourceInitializationError: unable to pull secrets or registry auth: execution resource
retrieval failed: unable to retrieve secret from asm: service call has been retried 5 time(s):
failed to fetch secret arn:aws:secretsmanager:eu-west-2:123456789012:secret:prod/db-AbCdEf
from secrets manager: … i/o timeout The middle tells you which service: asm is Secrets Manager, ssm is Parameter Store, and a registry-auth variant names ECR. The ending tells you why: a timeout means networking, AccessDeniedException means permissions, ResourceNotFoundException means the reference is wrong.
Causes and fixes
| Ending | Cause | Fix |
|---|---|---|
i/o timeout / context deadline exceeded | Task subnets can't reach the service endpoint | NAT gateway, public IP for Fargate in public subnets, or VPC endpoints |
AccessDeniedException | Execution role lacks secretsmanager:GetSecretValue or ssm:GetParameters | Grant them on the specific ARNs |
| KMS access denied | Secret encrypted with a customer managed key the execution role can't use | Grant kms:Decrypt on that key |
ResourceNotFoundException | Wrong ARN, wrong region, or a secret that was deleted | Use the full ARN from describe-secret |
| Invalid JSON key or version | Bad valueFrom suffix for a JSON key | Use arn…:secret:prod/db-AbCdEf:password:: |
The execution role is the one in executionRoleArn; the task role (taskRoleArn) isn't used for this step. Secrets are fetched once when the task starts, so a rotated secret only reaches the container when the task is replaced.
Check the task definition
aws ecs describe-task-definition --task-definition api \
--query 'taskDefinition.{exec: executionRoleArn, secrets: containerDefinitions[].secrets[]}' Example of correct references for a whole secret, one JSON key from a secret, and a Parameter Store parameter:
"secrets": [
{ "name": "DB_SECRET", "valueFrom": "arn:aws:secretsmanager:eu-west-2:123456789012:secret:prod/db-AbCdEf" },
{ "name": "DB_PASSWORD", "valueFrom": "arn:aws:secretsmanager:eu-west-2:123456789012:secret:prod/db-AbCdEf:password::" },
{ "name": "API_URL", "valueFrom": "arn:aws:ssm:eu-west-2:123456789012:parameter/prod/api-url" }
] A matching execution role policy:
{ "Effect": "Allow",
"Action": ["secretsmanager:GetSecretValue", "ssm:GetParameters"],
"Resource": ["arn:aws:secretsmanager:eu-west-2:123456789012:secret:prod/db-*",
"arn:aws:ssm:eu-west-2:123456789012:parameter/prod/*"] }Networking for private subnets
Without NAT, the task's subnets need interface endpoints for each service the task definition uses: secretsmanager, ssm, ecr.api, ecr.dkr, logs, plus the s3 gateway endpoint for image layers, and kms if a customer managed key is involved. Their security groups must accept HTTPS from the tasks.
Other ResourceInitializationError variants have the same flavour: failed to validate logger args usually means the awslogs log group doesn't exist (create it, or set awslogs-create-group with permission to create it), and EFS mount failures point at mount targets or security groups for NFS.
Questions
What is ResourceInitializationError in ECS?
ECS couldn't prepare resources a task needs before its containers start, such as secrets, the ECR auth token, the log group or EFS volumes. The task stops without running your code.
Which IAM role needs access to secrets in ECS?
The task execution role. It needs secretsmanager:GetSecretValue or ssm:GetParameters on the referenced ARNs, and kms:Decrypt if a customer managed key encrypts them.
How do I reference a JSON key from Secrets Manager in ECS?
Append the key to the secret ARN in valueFrom, in the form arn:…:secret:name-AbCdEf:jsonKey:: with empty version stage and version ID.

This error tends to start after a networking or secrets change, and it stops every new task from launching. CloudConsole, a phone app for AWS, is built to show service events and alarms on your phone so you notice before running tasks are replaced.
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