[Solved] CodeBuild "Internal Service Error" - Fix Region-Wide Provisioning Failures

[Solved] CodeBuild "Internal Service Error" - Provisioning Failures Across an Entire Region

The error "Internal Service Error: CodeBuild is experiencing issues" can show up even when nothing in your own pipeline configuration has changed. This guide walks through how we diagnosed a region-wide CodeBuild failure and what actually fixed it.

Internal Service Error: CodeBuild is experiencing issues

What Causes This Error?

This error typically occurs when:

• A source credential (Bitbucket/GitHub/GitLab) connected to your account has become corrupted
• The credential was re-imported or re-authorized multiple times in a short window
• CodeBuild's provisioning step fails while validating account-level source connections
• The issue is regional rather than tied to any single project

Step 1 - Rule Out Your Own Configuration First

Before assuming an AWS-side fault, check the usual suspects:

# Check recent changes to compute fleets
aws codebuild list-compute-fleets --region us-east-2

# Check recent changes to build images / VPC config on a project
aws codebuild batch-get-projects --names <project-name> --region us-east-2

If none of these have changed and the failure is affecting multiple unrelated projects at the same time, the issue is likely outside your project configuration.

Step 2 - Isolate With a Minimal Test Project

Create a brand-new CodeBuild project with no custom VPC, no custom image, and no dependencies — just the defaults:

aws codebuild create-project \
  --name test-region-diagnostic \
  --source type=NO_SOURCE \
  --artifacts type=NO_ARTIFACTS \
  --environment type=LINUX_CONTAINER,image=aws/codebuild/standard:7.0,computeType=BUILD_GENERAL1_SMALL \
  --service-role <role-arn> \
  --region us-east-2

If this minimal project fails with the same error, run the identical test in a different region:

aws codebuild create-project \
  --name test-region-diagnostic \
  --source type=NO_SOURCE \
  --artifacts type=NO_ARTIFACTS \
  --environment type=LINUX_CONTAINER,image=aws/codebuild/standard:7.0,computeType=BUILD_GENERAL1_SMALL \
  --service-role <role-arn> \
  --region us-east-1

If the same project succeeds in one region and fails in another, the problem is scoped to the account in that specific region — not your pipeline.

Step 3 - Check Source Credential History

Region-wide CodeBuild provisioning failures are often traced back to a corrupted source credential. Check how many times it's been re-imported recently:

aws codebuild list-source-credentials --region us-east-2

If a credential has been re-imported several times in a short window (especially across different auth types — OAuth, access token, basic auth), that repeated re-import can corrupt its internal storage and break provisioning for every project in the region, even ones that don't use that credential.

Step 4 - Fix: Delete and Re-Save the Credential

1. Open the CodeBuild console
2. Go to Settings → Source credentials
3. Delete the corrupted credential
4. Re-save a freshly generated access token for the same source provider

No redeploys or infrastructure changes are needed — this is purely an account-level credential fix.

Step 5 - Confirm the Fix

Check the FailedBuilds metric after the fix to confirm builds are succeeding again:

aws cloudwatch get-metric-statistics \
  --namespace AWS/CodeBuild \
  --metric-name FailedBuilds \
  --region us-east-2 \
  --start-time 2026-07-10T00:00:00Z \
  --end-time 2026-07-11T00:00:00Z \
  --period 3600 \
  --statistics Sum

A flat zero across the period confirms the region has recovered.

Quick Summary

Step	Command / Action
Check project config	aws codebuild batch-get-projects --names <project>
Create minimal test	aws codebuild create-project --source type=NO_SOURCE ...
Compare across regions	Run the same test project in a second region
Check credential history	aws codebuild list-source-credentials
Fix	Console → Settings → Source credentials → delete & re-save token
Verify	aws cloudwatch get-metric-statistics --metric-name FailedBuilds

If you're seeing this error and none of your own configuration has changed, check your account's source credential import history before assuming it's a broader AWS outage — a single corrupted credential can silently break every CodeBuild project in a region.