AWS Batch compute environments auto-revert to 0 vCPUs in eu-west-2 - Service Issue?

Question

**PROBLEM SUMMARY:**
All AWS Batch compute environments in eu-west-2 are automatically reverting desiredvCpus to 0 within minutes, making it impossible to run any jobs. This affects both GPU and CPU instance types.

**REGION:** eu-west-2  
**BUSINESS IMPACT:** Production pipeline completely blocked

**REPRODUCE STEPS:**
1. Create any compute environment (tested: g5.xlarge, m5.large)
2. UpdateComputeEnvironment with desiredvCpus=4  
3. Wait 2-5 minutes
4. DesiredvCpus automatically reverts to 0
5. Auto Scaling Groups created with Max=0 (abnormal)

**TROUBLESHOOTING COMPLETED:**
- Tested 3 different compute environments: all same behavior
- Multiple instance types affected
- Service quotas verified OK
- IAM/networking validated
- AMI functionality confirmed
- No automation/Lambda interference found

**QUESTION FOR AWS:**
Is there a known service issue with AWS Batch in eu-west-2? This appears to be systematic failure affecting all compute environments regardless of configuration.

**SIMILAR ISSUES?**
Has anyone else experienced this behavior? All jobs remain RUNNABLE indefinitely because no instances can launch.

**WORKAROUND NEEDED:**
This is blocking production workload. Are there any known workarounds or alternative regions recommended?

Kidd Ip · Answer

How about this as workaround:

* Manually Set vCPUs – Try setting Min vCPUs to a non-zero value to prevent automatic scaling to zero.
* Test in Another Region – If possible, deploy a test environment in eu-west-1 or us-east-1 to see if the issue persists.

Joel Herd - AWS Batch · Answer

Hi, thank you for using Batch!

> All jobs remain RUNNABLE indefinitely because no instances can launch.

I recommend configuring notifications on blocked job queues (https://aws.amazon.com/blogs/hpc/introducing-new-alerts-to-help-users-detect-and-react-to-blocked-job-queues-in-aws-batch/). You will find out exactly the reason instances cannot launch.

> UpdateComputeEnvironment with desiredvCpus=4

If your goal is to persist compute, you should use min vCpu instead of the desired vCpu (https://docs.aws.amazon.com/batch/latest/userguide/updating-compute-environments.html). Batch will scale down compute if it is incompatible for incoming jobs to save customer costs. Desired vCpu dropping to 0 is expected in your case.

AWS Batch compute environments auto-revert to 0 vCPUs in eu-west-2 - Service Issue?

Add your answer

Relevant content