Bug Report: GCP Label Validation Error for Owner Field

[jimmycgz]

Bug Report: GCP Label Validation Error for Owner Field

Issue Summary

PerfKitBenchmarker fails to create GCP VMs when the user's email address (used as the owner label) contains periods (.), which violates GCP's label value requirements.

Error Message

ERROR: (gcloud.compute.instances.create) argument --labels: Bad value [firstname.lastname]: 
Only hyphens (-), underscores (_), lowercase characters, and numbers are allowed. 
International characters are allowed.

Root Cause

In perfkitbenchmarker/benchmark_spec.py, the _GetResourceDict() method sets the owner label directly from FLAGS.owner without sanitization:

tags = {
    resource_type.TIMEOUT_METADATA_KEY: timeout_utc.strftime(time_format),
    'create_time_utc': now_utc.strftime(time_format),
    'benchmark': self.name,
    'perfkit_uuid': self.uuid,
    'owner': FLAGS.owner,  # ❌ Not sanitized
    'benchmark_uid': self.uid,
}

The FLAGS.owner value is derived from the user's GCP account email (e.g., firstname.lastname@example.com), which when extracted becomes firstname.lastname. The period character violates GCP's label requirements.

GCP Label Requirements

According to GCP documentation:

• Label values can only contain lowercase letters, numeric characters, underscores, and hyphens
• Maximum length: 63 characters
• Cannot start or end with underscores

Impact

• Severity: High - Blocks VM creation for users with periods in their email usernames
• Affected Users: Any user whose GCP account email contains periods (e.g., firstname.lastname@domain.com)
• Workaround: None available without code modification

Proposed Solution

Apply the existing _SafeLabelKeyOrValue() sanitization method to the owner field:

tags = {
    resource_type.TIMEOUT_METADATA_KEY: timeout_utc.strftime(time_format),
    'create_time_utc': now_utc.strftime(time_format),
    'benchmark': self.name,
    'perfkit_uuid': self.uuid,
    'owner': self._SafeLabelKeyOrValue(FLAGS.owner),  # ✅ Sanitized
    'benchmark_uid': self.uid,
}

This method already exists in the same class and is used for metadata key-value pairs. It:

1. Converts to lowercase
2. Replaces invalid characters with underscores
3. Truncates to 63 characters
4. Strips leading/trailing underscores

Example Transformation

• Input: firstname.lastname@example.com → firstname.lastname
• Output: firstname_lastname (GCP-compliant)

File Location

File: perfkitbenchmarker/benchmark_spec.py
Method: _GetResourceDict()
Line: ~1398 (may vary by version)

Reproduction Steps

1. Use a GCP account with a period in the username (e.g., firstname.lastname@domain.com)
2. Run any PKB benchmark that creates GCP VMs
3. Observe the label validation error during VM creation

# 1. Clone the repository
git clone <repository-url>
cd PerfKitBenchmarker

# 2. Create Python virtual environment (first time only)
python3 -m venv ../venv_nginx

# 3. Activate virtual environment
source ../venv_nginx/bin/activate
## python3 --version
## 3.13

# 4. Install Python dependencies
pip install --upgrade pip
pip install -r requirements.txt
pip install -r requirements-testing.txt

# 5. Authenticate with GCP
gcloud auth login
gcloud auth application-default login

# 6. Set your GCP project
gcloud config set project your-gcp-project

# First time test
python3 ./pkb.py

Error output

STDERR: 
2025-11-17 11:38:49,617 22b9dbcb Thread-8 (_ExecuteBackgroundThreadTasks) cluster_boot(1/1) INFO     Running: gcloud compute instances create pkb-22b9dbcb-0 --async --boot-disk-auto-delete --format json --image-family ubuntu-2404-lts-amd64 --image-project ubuntu-os-cloud --labels benchmark=cluster_boot,benchmark_uid=cluster_boot0,create_time_utc=20251117t163849z,owner=first.last,perfkit_uuid=22b9dbcb-ad6f6829-7d7d-4512-9e59-48a52aa12517,timeout_utc=20251117t203849z --machine-type n1-standard-2 --metadata benchmark=cluster_boot,benchmark_uid=cluster_boot0,create_time_utc=20251117t163849z,enable-oslogin=FALSE,owner=first.last,perfkit_uuid=22b9dbcb-ad6f6829-7d7d-4512-9e59-48a52aa12517,timeout_utc=20251117t203849z,vm_nature=ephemeral --metadata-from-file sshKeys=/tmp/perfkitbenchmarker/runs/22b9dbcb/key-metadatar48_e75r --no-restart-on-failure --project your-gcp-project --quiet --tags perfkitbenchmarker --zone us-central1-a --network-interface network=pkb-network-22b9dbcb,nic-type=GVNIC,network-tier=PREMIUM
2025-11-17 11:38:50,295 22b9dbcb Thread-8 (_ExecuteBackgroundThreadTasks) cluster_boot(1/1) INFO     Ran: {gcloud compute instances create pkb-22b9dbcb-0 --async --boot-disk-auto-delete --format json --image-family ubuntu-2404-lts-amd64 --image-project ubuntu-os-cloud --labels benchmark=cluster_boot,benchmark_uid=cluster_boot0,create_time_utc=20251117t163849z,owner=first.last,perfkit_uuid=22b9dbcb-ad6f6829-7d7d-4512-9e59-48a52aa12517,timeout_utc=20251117t203849z --machine-type n1-standard-2 --metadata benchmark=cluster_boot,benchmark_uid=cluster_boot0,create_time_utc=20251117t163849z,enable-oslogin=FALSE,owner=first.last,perfkit_uuid=22b9dbcb-ad6f6829-7d7d-4512-9e59-48a52aa12517,timeout_utc=20251117t203849z,vm_nature=ephemeral --metadata-from-file sshKeys=/tmp/perfkitbenchmarker/runs/22b9dbcb/key-metadatar48_e75r --no-restart-on-failure --project your-gcp-project --quiet --tags perfkitbenchmarker --zone us-central1-a --network-interface network=pkb-network-22b9dbcb,nic-type=GVNIC,network-tier=PREMIUM}
ReturnCode:2
STDOUT: 
STDERR: ERROR: (gcloud.compute.instances.create) argument --labels: Bad value [first.last]: Only hyphens (-), underscores (_), lowercase characters, and numbers are allowed. International characters are allowed.
Usage: gcloud compute instances create INSTANCE_NAMES [INSTANCE_NAMES ...] [optional flags]
  optional flags may be  --accelerator | --address | --no-address | --async |
                         --availability-domain | --boot-disk-auto-delete |
                         --boot-disk-device-name | --boot-disk-interface |
                         --boot-disk-kms-key | --boot-disk-kms-keyring |
       
  ...

Testing After applying the fix:

cd PerfKitBenchmarker
python3 ./pkb.py

...
2025-11-17 12:28:55,253 611bd312 MainThread INFO     Benchmark run statuses:
--------------------------------------------------------
Name          UID            Status     Failed Substatus
--------------------------------------------------------
cluster_boot  cluster_boot0  SUCCEEDED                  
--------------------------------------------------------
Success rate: 100.00% (1/1)
2025-11-17 12:28:55,253 611bd312 MainThread INFO     Complete logs can be found at: /tmp/perfkitbenchmarker/runs/611bd312/pkb.log
2025-11-17 12:28:55,253 611bd312 MainThread INFO     Completion statuses can be found at: /tmp/perfkitbenchmarker/runs/611bd312/completion_statuses.json
2025-11-17 12:28:55,253 611bd312 MainThread INFO     PKB exiting with return_code 0

Expected: VM creation succeeds with sanitized owner label

Additional Context

• The _SafeLabelKeyOrValue() method is already used for custom metadata labels but was not applied to core tags like owner
• This is a consistency issue - all label values should be sanitized, not just user-provided metadata
• The fix is minimal and uses existing, tested code

Related Code

The _SafeLabelKeyOrValue() method (lines 1383-1393):

def _SafeLabelKeyOrValue(self, key):
    result = ''.join(
        c if self._IsSafeKeyOrValueCharacter(c) else '_' for c in key.lower()
    )
    
    # max length constraints on keys and values
    # https://cloud.google.com/resource-manager/docs/creating-managing-labels
    max_safe_length = 63
    # GCP labels are not allowed to start or end with '_'
    return result[:max_safe_length].strip('_')

Version Information

• PerfKitBenchmarker Version: v1.12.0-5931-g71964578
• Python Version: 3.13
• GCP SDK: Latest
• OS: macOS