Source code for spack.cmd.ci

# Copyright Spack Project Developers. See COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)

import json
import os
import shutil
from urllib.parse import urlparse, urlunparse

import llnl.util.filesystem as fs
import llnl.util.tty as tty
import llnl.util.tty.color as clr

import spack.binary_distribution as bindist
import spack.ci as spack_ci
import spack.cmd
import spack.cmd.buildcache as buildcache
import spack.cmd.common.arguments
import spack.config as cfg
import spack.environment as ev
import spack.hash_types as ht
import spack.mirrors.mirror
import spack.util.gpg as gpg_util
import spack.util.timer as timer
import spack.util.url as url_util
import spack.util.web as web_util

description = "manage continuous integration pipelines"
section = "build"
level = "long"

SPACK_COMMAND = "spack"
INSTALL_FAIL_CODE = 1
FAILED_CREATE_BUILDCACHE_CODE = 100


[docs] def deindent(desc): return desc.replace(" ", "")
[docs] def unicode_escape(path: str) -> str: """Returns transformed path with any unicode characters replaced with their corresponding escapes""" return path.encode("unicode-escape").decode("utf-8")
[docs] def setup_parser(subparser): setup_parser.parser = subparser subparsers = subparser.add_subparsers(help="CI sub-commands") # Dynamic generation of the jobs yaml from a spack environment generate = subparsers.add_parser( "generate", description=deindent(ci_generate.__doc__), help=spack.cmd.first_line(ci_generate.__doc__), ) generate.add_argument( "--output-file", default=None, help="pathname for the generated gitlab ci yaml file\n\n" "path to the file where generated jobs file should be written. " "default is .gitlab-ci.yml in the root of the repository", ) prune_dag_group = generate.add_mutually_exclusive_group() prune_dag_group.add_argument( "--prune-dag", action="store_true", dest="prune_dag", default=True, help="skip up-to-date specs\n\n" "do not generate jobs for specs that are up-to-date on the mirror", ) prune_dag_group.add_argument( "--no-prune-dag", action="store_false", dest="prune_dag", default=True, help="process up-to-date specs\n\n" "generate jobs for specs even when they are up-to-date on the mirror", ) prune_ext_group = generate.add_mutually_exclusive_group() prune_ext_group.add_argument( "--prune-externals", action="store_true", dest="prune_externals", default=True, help="skip external specs\n\n" "do not generate jobs for specs that are marked as external", ) prune_ext_group.add_argument( "--no-prune-externals", action="store_false", dest="prune_externals", default=True, help="process external specs\n\n" "generate jobs for specs even when they are marked as external", ) generate.add_argument( "--check-index-only", action="store_true", dest="index_only", default=False, help="only check spec state from buildcache indices\n\n" "Spack always checks specs against configured binary mirrors, regardless of the DAG " "pruning option. if enabled, Spack will assume all remote buildcache indices are " "up-to-date when assessing whether the spec on the mirror, if present, is up-to-date. " "this has the benefit of reducing pipeline generation time but at the potential cost of " "needlessly rebuilding specs when the indices are outdated. if not enabled, Spack will " "fetch remote spec files directly to assess whether the spec on the mirror is up-to-date", ) generate.add_argument( "--artifacts-root", default="jobs_scratch_dir", help="path to the root of the artifacts directory\n\n" "The spack ci module assumes it will normally be run from within your project " "directory, wherever that is checked out to run your ci. The artifacts root directory " "should specifiy a name that can safely be used for artifacts within your project " "directory.", ) generate.set_defaults(func=ci_generate) spack.cmd.common.arguments.add_concretizer_args(generate) spack.cmd.common.arguments.add_common_arguments(generate, ["jobs"]) # Rebuild the buildcache index associated with the mirror in the # active, gitlab-enabled environment. index = subparsers.add_parser( "rebuild-index", description=deindent(ci_reindex.__doc__), help=spack.cmd.first_line(ci_reindex.__doc__), ) index.set_defaults(func=ci_reindex) # Handle steps of a ci build/rebuild rebuild = subparsers.add_parser( "rebuild", description=deindent(ci_rebuild.__doc__), help=spack.cmd.first_line(ci_rebuild.__doc__), ) rebuild.add_argument( "-t", "--tests", action="store_true", default=False, help="run stand-alone tests after the build", ) rebuild.add_argument( "--fail-fast", action="store_true", default=False, help="stop stand-alone tests after the first failure", ) rebuild.set_defaults(func=ci_rebuild) spack.cmd.common.arguments.add_common_arguments(rebuild, ["jobs"]) # Facilitate reproduction of a failed CI build job reproduce = subparsers.add_parser( "reproduce-build", description=deindent(ci_reproduce.__doc__), help=spack.cmd.first_line(ci_reproduce.__doc__), ) reproduce.add_argument( "job_url", help="URL of GitLab job web page or artifact", type=_gitlab_artifacts_url ) reproduce.add_argument( "--runtime", help="Container runtime to use.", default="docker", choices=["docker", "podman"], ) reproduce.add_argument( "--working-dir", help="where to unpack artifacts", default=os.path.join(os.getcwd(), "ci_reproduction"), ) reproduce.add_argument( "-s", "--autostart", help="Run docker reproducer automatically", action="store_true" ) gpg_group = reproduce.add_mutually_exclusive_group(required=False) gpg_group.add_argument( "--gpg-file", help="Path to public GPG key for validating binary cache installs" ) gpg_group.add_argument( "--gpg-url", help="URL to public GPG key for validating binary cache installs" ) reproduce.set_defaults(func=ci_reproduce)
[docs] def ci_generate(args): """generate jobs file from a CI-aware spack file if you want to report the results on CDash, you will need to set the SPACK_CDASH_AUTH_TOKEN before invoking this command. the value must be the CDash authorization token needed to create a build group and register all generated jobs under it """ env = spack.cmd.require_active_env(cmd_name="ci generate") spack_ci.generate_pipeline(env, args)
[docs] def ci_reindex(args): """rebuild the buildcache index for the remote mirror use the active, gitlab-enabled environment to rebuild the buildcache index for the associated mirror """ env = spack.cmd.require_active_env(cmd_name="ci rebuild-index") yaml_root = env.manifest[ev.TOP_LEVEL_KEY] if "mirrors" not in yaml_root or len(yaml_root["mirrors"].values()) < 1: tty.die("spack ci rebuild-index requires an env containing a mirror") ci_mirrors = yaml_root["mirrors"] mirror_urls = [url for url in ci_mirrors.values()] remote_mirror_url = mirror_urls[0] mirror = spack.mirrors.mirror.Mirror(remote_mirror_url) buildcache.update_index(mirror, update_keys=True)
[docs] def ci_rebuild(args): """rebuild a spec if it is not on the remote mirror check a single spec against the remote mirror, and rebuild it from source if the mirror does not contain the hash """ rebuild_timer = timer.Timer() env = spack.cmd.require_active_env(cmd_name="ci rebuild") # Make sure the environment is "gitlab-enabled", or else there's nothing # to do. ci_config = cfg.get("ci") if not ci_config: tty.die("spack ci rebuild requires an env containing ci cfg") # Grab the environment variables we need. These either come from the # pipeline generation step ("spack ci generate"), where they were written # out as variables, or else provided by GitLab itself. pipeline_artifacts_dir = os.environ.get("SPACK_ARTIFACTS_ROOT") job_log_dir = os.environ.get("SPACK_JOB_LOG_DIR") job_test_dir = os.environ.get("SPACK_JOB_TEST_DIR") repro_dir = os.environ.get("SPACK_JOB_REPRO_DIR") concrete_env_dir = os.environ.get("SPACK_CONCRETE_ENV_DIR") ci_job_name = os.environ.get("CI_JOB_NAME") signing_key = os.environ.get("SPACK_SIGNING_KEY") job_spec_pkg_name = os.environ.get("SPACK_JOB_SPEC_PKG_NAME") job_spec_dag_hash = os.environ.get("SPACK_JOB_SPEC_DAG_HASH") spack_pipeline_type = os.environ.get("SPACK_PIPELINE_TYPE") spack_ci_stack_name = os.environ.get("SPACK_CI_STACK_NAME") rebuild_everything = os.environ.get("SPACK_REBUILD_EVERYTHING") require_signing = os.environ.get("SPACK_REQUIRE_SIGNING") # If signing key was provided via "SPACK_SIGNING_KEY", then try to import it. if signing_key: spack_ci.import_signing_key(signing_key) # Fail early if signing is required but we don't have a signing key sign_binaries = require_signing is not None and require_signing.lower() == "true" if sign_binaries and not spack_ci.can_sign_binaries(): gpg_util.list(False, True) tty.die("SPACK_REQUIRE_SIGNING=True => spack must have exactly one signing key") # Construct absolute paths relative to current $CI_PROJECT_DIR ci_project_dir = os.environ.get("CI_PROJECT_DIR") pipeline_artifacts_dir = os.path.join(ci_project_dir, pipeline_artifacts_dir) job_log_dir = os.path.join(ci_project_dir, job_log_dir) job_test_dir = os.path.join(ci_project_dir, job_test_dir) repro_dir = os.path.join(ci_project_dir, repro_dir) concrete_env_dir = os.path.join(ci_project_dir, concrete_env_dir) # Debug print some of the key environment variables we should have received tty.debug("pipeline_artifacts_dir = {0}".format(pipeline_artifacts_dir)) tty.debug("job_spec_pkg_name = {0}".format(job_spec_pkg_name)) # Query the environment manifest to find out whether we're reporting to a # CDash instance, and if so, gather some information from the manifest to # support that task. cdash_config = cfg.get("cdash") cdash_handler = None if "build-group" in cdash_config: cdash_handler = spack_ci.CDashHandler(cdash_config) tty.debug("cdash url = {0}".format(cdash_handler.url)) tty.debug("cdash project = {0}".format(cdash_handler.project)) tty.debug("cdash project_enc = {0}".format(cdash_handler.project_enc)) tty.debug("cdash build_name = {0}".format(cdash_handler.build_name)) tty.debug("cdash build_stamp = {0}".format(cdash_handler.build_stamp)) tty.debug("cdash site = {0}".format(cdash_handler.site)) tty.debug("cdash build_group = {0}".format(cdash_handler.build_group)) # Is this a pipeline run on a spack PR or a merge to develop? It might # be neither, e.g. a pipeline run on some environment repository. spack_is_pr_pipeline = spack_pipeline_type == "spack_pull_request" spack_is_develop_pipeline = spack_pipeline_type == "spack_protected_branch" tty.debug( "Pipeline type - PR: {0}, develop: {1}".format( spack_is_pr_pipeline, spack_is_develop_pipeline ) ) full_rebuild = True if rebuild_everything and rebuild_everything.lower() == "true" else False pipeline_mirrors = spack.mirrors.mirror.MirrorCollection(binary=True) buildcache_destination = None if "buildcache-destination" not in pipeline_mirrors: tty.die("spack ci rebuild requires a mirror named 'buildcache-destination") buildcache_destination = pipeline_mirrors["buildcache-destination"] # Get the concrete spec to be built by this job. try: job_spec = env.get_one_by_hash(job_spec_dag_hash) except AssertionError: tty.die("Could not find environment spec with hash {0}".format(job_spec_dag_hash)) job_spec_json_file = "{0}.json".format(job_spec_pkg_name) job_spec_json_path = os.path.join(repro_dir, job_spec_json_file) # To provide logs, cdash reports, etc for developer download/perusal, # these things have to be put into artifacts. This means downstream # jobs that "need" this job will get those artifacts too. So here we # need to clean out the artifacts we may have got from upstream jobs. cdash_report_dir = os.path.join(pipeline_artifacts_dir, "cdash_report") if os.path.exists(cdash_report_dir): shutil.rmtree(cdash_report_dir) if os.path.exists(job_log_dir): shutil.rmtree(job_log_dir) if os.path.exists(job_test_dir): shutil.rmtree(job_test_dir) if os.path.exists(repro_dir): shutil.rmtree(repro_dir) # Now that we removed them if they existed, create the directories we # need for storing artifacts. The cdash_report directory will be # created internally if needed. os.makedirs(job_log_dir) os.makedirs(job_test_dir) os.makedirs(repro_dir) # Copy the concrete environment files to the repro directory so we can # expose them as artifacts and not conflict with the concrete environment # files we got as artifacts from the upstream pipeline generation job. # Try to cast a slightly wider net too, and hopefully get the generated # pipeline yaml. If we miss it, the user will still be able to go to the # pipeline generation job and get it from there. target_dirs = [concrete_env_dir, pipeline_artifacts_dir] for dir_to_list in target_dirs: for file_name in os.listdir(dir_to_list): src_file = os.path.join(dir_to_list, file_name) if os.path.isfile(src_file): dst_file = os.path.join(repro_dir, file_name) shutil.copyfile(src_file, dst_file) # Write this job's spec json into the reproduction directory, and it will # also be used in the generated "spack install" command to install the spec tty.debug("job concrete spec path: {0}".format(job_spec_json_path)) with open(job_spec_json_path, "w", encoding="utf-8") as fd: fd.write(job_spec.to_json(hash=ht.dag_hash)) # Write some other details to aid in reproduction into an artifact repro_file = os.path.join(repro_dir, "repro.json") repro_details = { "job_name": ci_job_name, "job_spec_json": job_spec_json_file, "ci_project_dir": ci_project_dir, } with open(repro_file, "w", encoding="utf-8") as fd: fd.write(json.dumps(repro_details)) # Write information about spack into an artifact in the repro dir spack_info = spack_ci.get_spack_info() spack_info_file = os.path.join(repro_dir, "spack_info.txt") with open(spack_info_file, "wb") as fd: fd.write(b"\n") fd.write(spack_info.encode("utf8")) fd.write(b"\n") matches = None if full_rebuild else bindist.get_mirrors_for_spec(job_spec, index_only=False) if matches: # Got a hash match on at least one configured mirror. All # matches represent the fully up-to-date spec, so should all be # equivalent. If artifacts mirror is enabled, we just pick one # of the matches and download the buildcache files from there to # the artifacts, so they're available to be used by dependent # jobs in subsequent stages. tty.msg("No need to rebuild {0}, found hash match at: ".format(job_spec_pkg_name)) for match in matches: tty.msg(" {0}".format(match["mirror_url"])) # Now we are done and successful return 0 # No hash match anywhere means we need to rebuild spec # Start with spack arguments spack_cmd = [SPACK_COMMAND, "--color=always", "--backtrace", "--verbose", "install"] config = cfg.get("config") if not config["verify_ssl"]: spack_cmd.append("-k") install_args = [ f'--use-buildcache={spack_ci.common.win_quote("package:never,dependencies:only")}' ] can_verify = spack_ci.can_verify_binaries() verify_binaries = can_verify and spack_is_pr_pipeline is False if not verify_binaries: install_args.append("--no-check-signature") if args.jobs: install_args.append(f"-j{args.jobs}") slash_hash = spack_ci.common.win_quote("/" + job_spec.dag_hash()) # Arguments when installing the root from sources deps_install_args = install_args + ["--only=dependencies"] root_install_args = install_args + ["--keep-stage", "--only=package"] if cdash_handler: # Add additional arguments to `spack install` for CDash reporting. root_install_args.extend(cdash_handler.args()) commands = [ # apparently there's a race when spack bootstraps? do it up front once [SPACK_COMMAND, "-e", unicode_escape(env.path), "bootstrap", "now"], spack_cmd + deps_install_args + [slash_hash], spack_cmd + root_install_args + [slash_hash], ] tty.debug("Installing {0} from source".format(job_spec.name)) install_exit_code = spack_ci.process_command("install", commands, repro_dir) # Now do the post-install tasks tty.debug("spack install exited {0}".format(install_exit_code)) # If a spec fails to build in a spack develop pipeline, we add it to a # list of known broken hashes. This allows spack PR pipelines to # avoid wasting compute cycles attempting to build those hashes. if install_exit_code == INSTALL_FAIL_CODE and spack_is_develop_pipeline: tty.debug("Install failed on develop") if "broken-specs-url" in ci_config: broken_specs_url = ci_config["broken-specs-url"] dev_fail_hash = job_spec.dag_hash() broken_spec_path = url_util.join(broken_specs_url, dev_fail_hash) tty.msg("Reporting broken develop build as: {0}".format(broken_spec_path)) spack_ci.write_broken_spec( broken_spec_path, job_spec_pkg_name, spack_ci_stack_name, os.environ.get("CI_JOB_URL"), os.environ.get("CI_PIPELINE_URL"), job_spec.to_dict(hash=ht.dag_hash), ) # We generated the "spack install ..." command to "--keep-stage", copy # any logs from the staging directory to artifacts now spack_ci.copy_stage_logs_to_artifacts(job_spec, job_log_dir) # If the installation succeeded and we're running stand-alone tests for # the package, run them and copy the output. Failures of any kind should # *not* terminate the build process or preclude creating the build cache. broken_tests = ( "broken-tests-packages" in ci_config and job_spec.name in ci_config["broken-tests-packages"] ) reports_dir = fs.join_path(os.getcwd(), "cdash_report") if args.tests and broken_tests: tty.warn("Unable to run stand-alone tests since listed in ci's 'broken-tests-packages'") if cdash_handler: msg = "Package is listed in ci's broken-tests-packages" cdash_handler.report_skipped(job_spec, reports_dir, reason=msg) cdash_handler.copy_test_results(reports_dir, job_test_dir) elif args.tests: if install_exit_code == 0: try: # First ensure we will use a reasonable test stage directory stage_root = os.path.dirname(str(job_spec.package.stage.path)) test_stage = fs.join_path(stage_root, "spack-standalone-tests") tty.debug("Configuring test_stage to {0}".format(test_stage)) config_test_path = "config:test_stage:{0}".format(test_stage) cfg.add(config_test_path, scope=cfg.default_modify_scope()) # Run the tests, resorting to junit results if not using cdash log_file = ( None if cdash_handler else fs.join_path(test_stage, "ci-test-results.xml") ) spack_ci.run_standalone_tests( cdash=cdash_handler, job_spec=job_spec, fail_fast=args.fail_fast, log_file=log_file, repro_dir=repro_dir, ) except Exception as err: # If there is any error, just print a warning. msg = "Error processing stand-alone tests: {0}".format(str(err)) tty.warn(msg) finally: # Copy the test log/results files spack_ci.copy_test_logs_to_artifacts(test_stage, job_test_dir) if cdash_handler: cdash_handler.copy_test_results(reports_dir, job_test_dir) elif log_file: spack_ci.copy_files_to_artifacts(log_file, job_test_dir) else: tty.warn("No recognized test results reporting option") else: tty.warn("Unable to run stand-alone tests due to unsuccessful installation") if cdash_handler: msg = "Failed to install the package" cdash_handler.report_skipped(job_spec, reports_dir, reason=msg) cdash_handler.copy_test_results(reports_dir, job_test_dir) if install_exit_code == 0: # If the install succeeded, push it to the buildcache destination. Failure to push # will result in a non-zero exit code. Pushing is best-effort. for result in spack_ci.create_buildcache( input_spec=job_spec, destination_mirror_urls=[buildcache_destination.push_url], sign_binaries=spack_ci.can_sign_binaries(), ): if not result.success: install_exit_code = FAILED_CREATE_BUILDCACHE_CODE (tty.msg if result.success else tty.error)( f'{"Pushed" if result.success else "Failed to push"} ' f'{job_spec.format("{name}{@version}{/hash:7}", color=clr.get_color_when())} ' f"to {result.url}" ) # If this is a develop pipeline, check if the spec that we just built is # on the broken-specs list. If so, remove it. if spack_is_develop_pipeline and "broken-specs-url" in ci_config: broken_specs_url = ci_config["broken-specs-url"] just_built_hash = job_spec.dag_hash() broken_spec_path = url_util.join(broken_specs_url, just_built_hash) if web_util.url_exists(broken_spec_path): tty.msg("Removing {0} from the list of broken specs".format(broken_spec_path)) try: web_util.remove_url(broken_spec_path) except Exception as err: # If there is an S3 error (e.g., access denied or connection # error), the first non boto-specific class in the exception # hierarchy is Exception. Just print a warning and return. msg = "Error removing {0} from broken specs list: {1}" tty.warn(msg.format(broken_spec_path, err)) else: # If the install did not succeed, print out some instructions on how to reproduce this # build failure outside of the pipeline environment. tty.debug("spack install exited non-zero, will not create buildcache") api_root_url = os.environ.get("CI_API_V4_URL") ci_project_id = os.environ.get("CI_PROJECT_ID") ci_job_id = os.environ.get("CI_JOB_ID") repro_job_url = f"{api_root_url}/projects/{ci_project_id}/jobs/{ci_job_id}/artifacts" # Control characters cause this to be printed in blue so it stands out print( f""" \033[34mTo reproduce this build locally, run: spack ci reproduce-build {repro_job_url} [--working-dir <dir>] [--autostart] If this project does not have public pipelines, you will need to first: export GITLAB_PRIVATE_TOKEN=<generated_token> ... then follow the printed instructions.\033[0;0m """ ) rebuild_timer.stop() try: with open("install_timers.json", "w", encoding="utf-8") as timelog: extra_attributes = {"name": ".ci-rebuild"} rebuild_timer.write_json(timelog, extra_attributes=extra_attributes) except Exception as e: tty.debug(str(e)) # Tie job success/failure to the success/failure of building the spec return install_exit_code
[docs] def ci_reproduce(args): """generate instructions for reproducing the spec rebuild job artifacts of the provided gitlab pipeline rebuild job's URL will be used to derive instructions for reproducing the build locally """ # Allow passing GPG key for reprocuding protected CI jobs if args.gpg_file: gpg_key_url = url_util.path_to_file_url(args.gpg_file) elif args.gpg_url: gpg_key_url = args.gpg_url else: gpg_key_url = None return spack_ci.reproduce_ci_job( args.job_url, args.working_dir, args.autostart, gpg_key_url, args.runtime )
def _gitlab_artifacts_url(url: str) -> str: """Take a URL either to the URL of the job in the GitLab UI, or to the artifacts zip file, and output the URL to the artifacts zip file.""" parsed = urlparse(url) if not parsed.scheme or not parsed.netloc: raise ValueError(url) parts = parsed.path.split("/") if len(parts) < 2: raise ValueError(url) # Just use API endpoints verbatim, they're probably generated by Spack. if parts[1] == "api": return url # If it's a URL to the job in the Gitlab UI, we may need to append the artifacts path. minus_idx = parts.index("-") # Remove repeated slashes in the remainder rest = [p for p in parts[minus_idx + 1 :] if p] # Now the format is jobs/X or jobs/X/artifacts/download if len(rest) < 2 or rest[0] != "jobs": raise ValueError(url) if len(rest) == 2: # replace jobs/X with jobs/X/artifacts/download rest.extend(("artifacts", "download")) # Replace the parts and unparse. parts[minus_idx + 1 :] = rest # Don't allow fragments / queries return urlunparse(parsed._replace(path="/".join(parts), fragment="", query=""))
[docs] def ci(parser, args): if args.func: return args.func(args)