Perf micro benchmark automation using cloudbuild

[jasha26]

Description

This PR introduces a robust Cloud Build configuration (cloudbuild/benchmarks-cloudbuild.yaml) to automate performance benchmarking for GCSFS. It handles the full lifecycle of the benchmark run: provisioning infrastructure (VMs and GCS buckets), setting up the environment, executing benchmark scripts, and uploading results.

Substitutions

The substitutions section defines variables that control the infrastructure setup and the specific benchmarks to be executed. These can be overridden at build time.

Variable	Description	Format / Example
_ZONE	The Google Cloud zone where the VM and Zonal buckets will be created.	us-central1-a
_INFRA_PREFIX	A prefix used for naming all generated resources (VMs, Buckets) to ensure uniqueness and easy identification.	gcsfs-perf
_VM_SERVICE_ACCOUNT	The email address of the Service Account attached to the benchmark VM.	service-account@project.iam.gserviceaccount.com
_BUCKET_TYPES	A space-separated list of bucket types to provision and test against.	zonal regional hns
_BENCHMARK_CONFIG	Defines the benchmark suites to run. Format: group1:config1,config2 group2:config1.	read:read_seq,read_rand write:write_seq

---

Key Cloud Build Steps

The pipeline is designed to be robust, ensuring resources are cleaned up even if tests fail.

1. Generate SSH Key (generate-ssh-key)

Creates a temporary SSH key pair (google_compute_engine) for this specific build run to securely communicate with the worker VM.

2. Initialize Variables (init-variables)

• Generates unique resource names (using the Build ID) for the VM and Buckets.
• Writes these variables to a shared file (build_vars.env) so subsequent steps can reference the exact resources created.

3. Create Buckets (create-buckets)

• Provisions the necessary GCS buckets (Regional, Zonal, HNS) in parallel based on _BUCKET_TYPES.
• Ensures a persistent Results Bucket exists to store benchmark data.

4. Create VM (create-vm)

• Provisions a high-performance Compute Engine VM (C4 instance with Hyperdisk) configured with the generated SSH keys and specific network settings.

5. Run Benchmarks (run-benchmarks)

• Wait for SSH: Polls the VM until it is ready to accept connections.
• Copy Source: Transfers the GCSFS source code to the VM using scp.
• Install Dependencies: Runs a remote script to install Python, pip, and required system libraries.
• Execute Tests: Iterates through the _BENCHMARK_CONFIG, running the Python benchmark script for each configuration against the provisioned buckets.
• Upload Results: Uploads generated JSON/CSV result files from the VM to the persistent GCS Results Bucket.
• Note: This step has allowFailure: true to ensure cleanup steps run even if benchmarks fail.

6. Cleanup (cleanup-ssh-key, cleanup-vm, delete-buckets)

• SSH Keys: Removes the temporary public key from the project's OS Login metadata.
• VM: Deletes the compute instance.
• Buckets: Deletes the temporary test buckets.

7. Check Failure (check-failure)

• Inspects a local marker file (/workspace/failure) created by the run-benchmarks step.
• If the marker exists, this step explicitly fails the build, ensuring the CI system reports a failure status after cleanup is complete.

Testing

• Verified via Cloud Build execution.
• Confirmed VM creation, dependency installation, and benchmark script execution.
• Verified results upload to GCS.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 50.00000% with 6 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@e0c0a9f). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
gcsfs/tests/conftest.py	50.00%	6 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #753   +/-   ##
=======================================
  Coverage        ?   75.37%           
=======================================
  Files           ?       18           
  Lines           ?     2761           
  Branches        ?        0           
=======================================
  Hits            ?     2081           
  Misses          ?      680           
  Partials        ?        0           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ankitaluthra1]

/gcbrun

[ankitaluthra1]

Lets remove test coverage related changes from this PR, there is another PR for test coverage which contains full changes, it would also need some repo changes, so lets not block this PR for it

[ankitaluthra1]

The output wont be accessible outside, should we document such caveats and add alternatives on how ppl can run this on their own

[jasha26]

The cloudbuild itself generates logs which prints the final results table. I think that is good enough output of you inspect logs after there.

If you mean the cav files in bucket, i think that's something that is already documented in PR description.

[ankitaluthra1]

/gcbrun