Add helper to run CIs for a branch using az pipelines. (#16843)

### Description
<!-- Describe your changes. -->
Add helper to run CIs for a branch using `az pipelines`.
This can be used to easily kick off multiple CIs for a branch prior to
creating a PR.

Update run_CIs_for_external_pr.py so the CI list can be shared. 
Request json output from `gh pr view` so the current state is more
easily parsed.

### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
This commit is contained in:
Scott McKay 2024-02-24 14:06:30 +10:00 committed by GitHub
parent 46c4d7fe4a
commit c12a20bef9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 185 additions and 59 deletions

View file

@ -0,0 +1,116 @@
#!/usr/bin/env python3
# Copyright (c) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License.
import argparse
import json
import os
import subprocess
import sys
import typing
from run_CIs_for_external_pr import get_pipeline_names
from util.platform_helpers import is_windows
def _parse_args():
parser = argparse.ArgumentParser(
os.path.basename(__file__),
formatter_class=argparse.RawDescriptionHelpFormatter,
description="""Run the CIs used to validate PRs for the specified branch.
If specified, the `--include` filter is applied first, followed by any `--exclude` filter.
Requires the Azure CLI with DevOps extension to be installed.
Azure CLI: https://learn.microsoft.com/en-us/cli/azure/install-azure-cli
DevOps extension: https://github.com/Azure/azure-devops-cli-extension
Configuration:
Login:`az login`
Configure ORT repo as default:
`az devops configure --defaults organization=https://dev.azure.com/onnxruntime project=onnxruntime`
Example usage:
List all CIs
`python run_CIs_for_branch.py --dry-run my/BranchName`
Run all CIs
`python run_CIs_for_branch.py my/BranchName`
Run only Linux CIs
`python run_CIs_for_branch.py --include linux my/BranchName`
Exclude training CIs
`python run_CIs_for_branch.py --exclude training my/BranchName`
Run non-training Linux CIs
`python run_CIs_for_branch.py --include linux --exclude training my/BranchName`
""",
)
parser.add_argument("-i", "--include", type=str, help="Include CIs that match this string. Case insensitive.")
parser.add_argument("-e", "--exclude", type=str, help="Exclude CIs that match this string. Case insensitive.")
parser.add_argument("--dry-run", action="store_true", help="Print selected CIs but do not run them.")
parser.add_argument("branch", type=str, help="Specify the branch to run.")
args = parser.parse_args()
return args
def _run_az_pipelines_command(command: typing.List[str]):
try:
az = "az.cmd" if is_windows() else "az"
az_output = subprocess.run([az, "pipelines", *command], capture_output=True, text=True, check=True)
except subprocess.CalledProcessError as cpe:
print(cpe)
print(cpe.stderr)
sys.exit(-1)
return az_output
def main():
args = _parse_args()
branch = args.branch
# To debug available pipelines:
# az_out = az_pipelines = _run_az_pipelines_command(["list"])
# pipeline_info = json.loads(az_out.stdout)
# print(pipeline_info)
pipelines = get_pipeline_names()
pipelines_to_run = []
if args.include:
value = args.include.lower().strip()
for p in pipelines:
if value in p.lower():
print(f"Including {p}")
pipelines_to_run.append(p)
else:
pipelines_to_run = pipelines
if args.exclude:
value = args.exclude.lower().strip()
cur_pipelines = pipelines_to_run
pipelines_to_run = []
for p in cur_pipelines:
if value in p.lower():
print(f"Excluding {p}")
else:
pipelines_to_run.append(p)
print("Pipelines to run:")
for p in pipelines_to_run:
print(f"\t{p}")
if args.dry_run:
sys.exit(0)
for pipeline in pipelines_to_run:
az_out = _run_az_pipelines_command(["run", "--branch", branch, "--name", pipeline])
run_output = json.loads(az_out.stdout)
if "id" in run_output:
build_url = f"https://dev.azure.com/onnxruntime/onnxruntime/_build/results?buildId={run_output['id']}"
print(f"{pipeline} build results: {build_url}&view=results")
else:
raise ValueError("Build id was not found in az output:\n" + run_output)
if __name__ == "__main__":
main()

View file

@ -3,71 +3,18 @@
# Licensed under the MIT License.
import argparse
import json
import os
import subprocess
import sys
import typing
def parse_args():
parser = argparse.ArgumentParser(
os.path.basename(__file__),
formatter_class=argparse.RawDescriptionHelpFormatter,
description="""Trigger CIs running for the specified pull request.
Requires the GitHub CLI to be installed. See https://github.com/cli/cli#installation for details.
After installation you will also need to setup an auth token to access the ONNX Runtime repository by running
`gh auth login`. Easiest is to run that from a directory in your local copy of the repo.
""",
)
parser.add_argument("pr", help="Specify the pull request ID.")
args = parser.parse_args()
return args
def run_gh_pr_command(command: typing.List[str], check=True):
try:
return subprocess.run(["gh", "pr", *command], capture_output=True, text=True, check=check)
except subprocess.CalledProcessError as cpe:
print(cpe)
print(cpe.stderr)
sys.exit(-1)
def main():
args = parse_args()
pr_id = args.pr
# validate PR
gh_out = run_gh_pr_command(["view", pr_id])
info = gh_out.stdout.split("\n")
for line in info:
pieces = line.split("\t")
if len(pieces) != 2:
continue
if pieces[0] == "state:":
if pieces[1] != "OPEN":
print(f"PR {pr_id} is not OPEN. Currently in state {pieces[1]}.")
sys.exit(-1)
print("Check passed pipelines")
gh_out = run_gh_pr_command(["checks", pr_id, "--required"], check=False)
# output format is a tab separated list of columns:
# (pipeline name) "\t" (status) "\t" (ran time) "\t" (url)
checked_pipelines = [
columns[0]
for columns in (line.strip().split("\t") for line in gh_out.stdout.split("\n"))
if len(columns) == 4 and columns[1] == "pass"
]
print("Adding azp run commands")
def get_pipeline_names():
# Current pipelines. These change semi-frequently and may need updating.
#
# Note: there is no easy way to get the list for azp "required" pipelines before they starts.
# we need to maintain this list manually.
#
# There is no easy way to get the list of "required" pipelines using `azp` before they are run,
# so we need to maintain this list manually.
# NOTE: This list is also used by run_CIs_for_branch.py
pipelines = [
# windows
"Windows ARM64 QNN CI Pipeline",
@ -91,7 +38,6 @@ def main():
"orttraining-linux-gpu-ci-pipeline",
"orttraining-ortmodule-distributed",
# checks
"onnxruntime-python-checks-ci-pipeline",
"onnxruntime-binary-size-checks-ci-pipeline",
# big models
"Big Models",
@ -101,14 +47,78 @@ def main():
"ONNX Runtime React Native CI Pipeline",
]
return pipelines
def _parse_args():
parser = argparse.ArgumentParser(
os.path.basename(__file__),
formatter_class=argparse.RawDescriptionHelpFormatter,
description="""Trigger CIs running for the specified pull request.
Requires the GitHub CLI to be installed. See https://github.com/cli/cli#installation for details.
After installation you will also need to setup an auth token to access the ONNX Runtime repository by running
`gh auth login`. Easiest is to run that from a directory in your local copy of the repo.
""",
)
parser.add_argument("pr", help="Specify the pull request ID.")
args = parser.parse_args()
return args
def run_gh_pr_command(command: typing.List[str], check: bool = True):
try:
return subprocess.run(["gh", "pr", *command], capture_output=True, text=True, check=check)
except subprocess.CalledProcessError as cpe:
print(cpe)
print(cpe.stderr)
sys.exit(-1)
def main():
args = _parse_args()
pr_id = args.pr
# validate PR
print("Checking PR is open")
gh_out = run_gh_pr_command(["view", "--json", "state", pr_id])
info = json.loads(gh_out.stdout)
if "state" not in info:
print(f"Could not get current state from `gh pr view` response of\n{gh_out.stdout}")
sys.exit(-1)
if info["state"] != "OPEN":
print(f"PR {pr_id} is not OPEN. Currently in state {info['state']}.")
sys.exit(0)
# This will return CIs that have run previously but not passed. We filter the CIs to run based on this, so it's
# fine for the initial response to have no info in it.
# `gh pr checks` exits with non-zero exit code when failures in pipeline exist, so we set `check` to False.
print("Checking for pipelines that have passed.")
gh_out = run_gh_pr_command(["checks", pr_id, "--required"], check=False)
# output format is a tab separated list of columns:
# (pipeline name) "\t" (status) "\t" (ran time) "\t" (url)
checked_pipelines = [
columns[0]
for columns in (line.strip().split("\t") for line in gh_out.stdout.split("\n"))
if len(columns) == 4 and columns[1] == "pass"
]
pipelines = get_pipeline_names()
# remove pipelines that have already run successfully
pipelines = [p for p in pipelines if p not in checked_pipelines]
print("Pipelines to run:")
for p in pipelines:
print("\t" + p)
# azp run is limited to 10 pipelines at a time
max_pipelines_per_comment = 10
start = 0
num_pipelines = len(pipelines)
print("Adding azp run commands")
while start < num_pipelines:
end = start + max_pipelines_per_comment
if end > num_pipelines: