release workflow: Add build scripts for jobs and means to upload pkgs

This commit is contained in:
Scott Wittenburg 2018-12-17 17:40:16 -07:00 committed by Peter Scheibel
parent a6e8e889b2
commit 5600c9f0d9
3 changed files with 541 additions and 0 deletions

13
bin/rebuild-index.sh Executable file
View file

@ -0,0 +1,13 @@
#!/bin/bash
# Copyright 2013-2019 Lawrence Livermore National Security, LLC and other
# Spack Project Developers. See the top-level COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)
set -x
SPACK_BIN_DIR="${CI_PROJECT_DIR}/bin"
export PATH="${SPACK_BIN_DIR}:${PATH}"
spack upload-s3 index

316
bin/rebuild-package.sh Executable file
View file

@ -0,0 +1,316 @@
#!/bin/bash
# Copyright 2013-2019 Lawrence Livermore National Security, LLC and other
# Spack Project Developers. See the top-level COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)
###
### This script represents a gitlab-ci job, corresponding to a single release
### spec. As such this script must first decide whether or not the spec it
### has been assigned is up to date on the remote binary mirror. If it is
### not (i.e. the source code has changed in a way that caused a change in the
### full_hash of the spec), this script will build the package, create a
### binary cache for it, and then push all related files to the remote binary
### mirror. This script also communicates with a remote CDash instance to
### share status on the package build process.
###
### The following environment variables are expected to be set in order for
### the various elements in this script to function properly. Listed first
### are two defaults we rely on from gitlab, then three we set up in the
### variables section of gitlab ourselves, and finally four variables
### written into the .gitlab-ci.yml file.
###
### CI_PROJECT_DIR
### CI_JOB_NAME
###
### AWS_ACCESS_KEY_ID
### AWS_SECRET_ACCESS_KEY
### SPACK_SIGNING_KEY
###
### CDASH_BASE_URL
### ROOT_SPEC
### DEPENDENCIES
### MIRROR_URL
###
shopt -s expand_aliases
export FORCE_UNSAFE_CONFIGURE=1
TEMP_DIR="${CI_PROJECT_DIR}/jobs_scratch_dir"
JOB_LOG_DIR="${TEMP_DIR}/logs"
SPEC_DIR="${TEMP_DIR}/specs"
LOCAL_MIRROR="${CI_PROJECT_DIR}/local_mirror"
BUILD_CACHE_DIR="${LOCAL_MIRROR}/build_cache"
SPACK_BIN_DIR="${CI_PROJECT_DIR}/bin"
CDASH_UPLOAD_URL="${CDASH_BASE_URL}/submit.php?project=Spack"
DEP_JOB_RELATEBUILDS_URL="${CDASH_BASE_URL}/api/v1/relateBuilds.php"
declare -a JOB_DEPS_PKG_NAMES
export SPACK_ROOT=${CI_PROJECT_DIR}
export PATH="${SPACK_BIN_DIR}:${PATH}"
export GNUPGHOME="${CI_PROJECT_DIR}/opt/spack/gpg"
mkdir -p ${JOB_LOG_DIR}
mkdir -p ${SPEC_DIR}
cleanup() {
set +x
if [ -z "$exit_code" ] ; then
exit_code=$1
if [ -z "$exit_code" ] ; then
exit_code=0
fi
restore_io
if [ "$( type -t finalize )" '=' 'function' ] ; then
finalize "$JOB_LOG_DIR/cdash_log.txt"
fi
# We can clean these out later on, once we have a good sense for
# how the logging infrastructure is working
# rm -rf "$JOB_LOG_DIR"
fi
\exit $exit_code
}
alias exit='cleanup'
begin_logging() {
trap "cleanup 1; \\exit \$exit_code" INT TERM QUIT
trap "cleanup 0; \\exit \$exit_code" EXIT
rm -rf "$JOB_LOG_DIR/cdash_log.txt"
# NOTE: Here, some redirects are set up
exec 3>&1 # fd 3 is now a dup of stdout
exec 4>&2 # fd 4 is now a dup of stderr
# stdout and stderr are joined and redirected to the log
exec &> "$JOB_LOG_DIR/cdash_log.txt"
set -x
}
restore_io() {
exec >&-
exec 2>&-
exec >&3
exec 2>&4
exec 3>&-
exec 4>&-
}
finalize() {
# If you define a finalize function:
# - it will always be called at the very end of the script
# - the log file will be passed in as the first argument, and
# - the code in this function will not be logged.
echo "The full log file is located at $1"
# TODO: send this log data to cdash!
}
check_error()
{
local last_exit_code=$1
local last_cmd=$2
if [[ ${last_exit_code} -ne 0 ]]; then
echo "${last_cmd} exited with code ${last_exit_code}"
echo "TERMINATING JOB"
exit 1
else
echo "${last_cmd} completed successfully"
fi
}
extract_build_id()
{
LINES_TO_SEARCH=$1
regex="buildSummary\.php\?buildid=([[:digit:]]+)"
SINGLE_LINE_OUTPUT=$(echo ${LINES_TO_SEARCH} | tr -d '\n')
if [[ ${SINGLE_LINE_OUTPUT} =~ ${regex} ]]; then
echo "${BASH_REMATCH[1]}"
else
echo "NONE"
fi
}
get_relate_builds_post_data()
{
cat <<EOF
{
"project": "${1}",
"buildid": ${2},
"relatedid": ${3},
"relationship": "depends on"
}
EOF
}
gen_full_specs_for_job_and_deps() {
read -ra PARTSARRAY <<< "${CI_JOB_NAME}"
local pkgName="${PARTSARRAY[0]}"
local pkgVersion="${PARTSARRAY[1]}"
local compiler="${PARTSARRAY[2]}"
local osarch="${PARTSARRAY[3]}"
JOB_SPEC_NAME="${pkgName}@${pkgVersion}%${compiler} arch=${osarch}"
JOB_PKG_NAME="${pkgName}"
SPEC_YAML_PATH="${SPEC_DIR}/${pkgName}.yaml"
local root_spec_name="${ROOT_SPEC}"
local spec_names_to_save="${pkgName}"
IFS=';' read -ra DEPS <<< "${DEPENDENCIES}"
for i in "${DEPS[@]}"; do
read -ra PARTSARRAY <<< "${i}"
pkgName="${PARTSARRAY[0]}"
spec_names_to_save="${spec_names_to_save} ${pkgName}"
JOB_DEPS_PKG_NAMES+=("${pkgName}")
done
spack -d buildcache save-yaml --specs "${spec_names_to_save}" --root-spec "${root_spec_name}" --yaml-dir "${SPEC_DIR}"
}
begin_logging
gen_full_specs_for_job_and_deps
echo "Building package ${JOB_SPEC_NAME}, ${HASH}, ${MIRROR_URL}"
# Finally, list the compilers spack knows about
echo "Compiler Configurations:"
spack config get compilers
# Make the build_cache directory if it doesn't exist
mkdir -p "${BUILD_CACHE_DIR}"
# Get buildcache name so we can write a CDash build id file in the right place.
# If we're unable to get the buildcache name, we may have encountered a problem
# concretizing the spec, or some other issue that will eventually cause the job
# to fail.
JOB_BUILD_CACHE_ENTRY_NAME=`spack -d buildcache get-buildcache-name --spec-yaml "${SPEC_YAML_PATH}"`
if [[ $? -ne 0 ]]; then
echo "ERROR, unable to get buildcache entry name for job ${CI_JOB_NAME} (spec: ${JOB_SPEC_NAME})"
exit 1
fi
# This should create the directory we referred to as GNUPGHOME earlier
spack gpg list
# Importing the secret key using gpg2 directly should allow to
# sign and verify both
set +x
KEY_IMPORT_RESULT=`echo ${SPACK_SIGNING_KEY} | base64 --decode | gpg2 --import`
check_error $? "gpg2 --import"
set -x
spack gpg list --trusted
spack gpg list --signing
# Whether we have to build the spec or download it pre-built, we expect to find
# the cdash build id file sitting in this location afterwards.
JOB_CDASH_ID_FILE="${BUILD_CACHE_DIR}/${JOB_BUILD_CACHE_ENTRY_NAME}.cdashid"
# Finally, we can check the spec we have been tasked with build against
# the built binary on the remote mirror to see if it needs to be rebuilt
spack -d buildcache check --spec-yaml "${SPEC_YAML_PATH}" --mirror-url "${MIRROR_URL}" --rebuild-on-error
if [[ $? -ne 0 ]]; then
# Configure mirror
spack mirror add local_artifact_mirror "file://${LOCAL_MIRROR}"
JOB_CDASH_ID="NONE"
# Install package, using the buildcache from the local mirror to
# satisfy dependencies.
BUILD_ID_LINE=`spack -d -k -v install --use-cache --cdash-upload-url "${CDASH_UPLOAD_URL}" --cdash-build "${JOB_SPEC_NAME}" --cdash-site "Spack AWS Gitlab Instance" --cdash-track "Experimental" -f "${SPEC_YAML_PATH}" | grep "buildSummary\\.php"`
check_error $? "spack install"
# By parsing the output of the "spack install" command, we can get the
# buildid generated for us by CDash
JOB_CDASH_ID=$(extract_build_id "${BUILD_ID_LINE}")
# Create buildcache entry for this package, reading the spec from the yaml
# file.
spack -d buildcache create --spec-yaml "${SPEC_YAML_PATH}" -a -f -d "${LOCAL_MIRROR}" --no-rebuild-index
check_error $? "spack buildcache create"
# Write the .cdashid file to the buildcache as well
echo "${JOB_CDASH_ID}" >> ${JOB_CDASH_ID_FILE}
# TODO: The upload-s3 command should eventually be replaced with something
# like: "spack buildcache put <mirror> <spec>", when that subcommand is
# properly implemented.
spack -d upload-s3 spec --base-dir "${LOCAL_MIRROR}" --spec-yaml "${SPEC_YAML_PATH}"
check_error $? "spack upload-s3 spec"
else
echo "spec ${JOB_SPEC_NAME} is already up to date on remote mirror, downloading it"
# Configure remote mirror so we can download buildcache entry
spack mirror add remote_binary_mirror ${MIRROR_URL}
# Now download it
spack -d buildcache download --spec-yaml "${SPEC_YAML_PATH}" --path "${BUILD_CACHE_DIR}/" --require-cdashid
check_error $? "spack buildcache download"
fi
# The next step is to relate this job to the jobs it depends on
if [ -f "${JOB_CDASH_ID_FILE}" ]; then
JOB_CDASH_BUILD_ID=$(<${JOB_CDASH_ID_FILE})
if [ "${JOB_CDASH_BUILD_ID}" == "NONE" ]; then
echo "ERROR: unable to read this jobs id from ${JOB_CDASH_ID_FILE}"
exit 1
fi
# Now get CDash ids for dependencies and "relate" each dependency build
# with this jobs build
for DEP_PKG_NAME in "${JOB_DEPS_PKG_NAMES[@]}"; do
echo "Getting cdash id for dependency --> ${DEP_PKG_NAME} <--"
DEP_SPEC_YAML_PATH="${SPEC_DIR}/${DEP_PKG_NAME}.yaml"
DEP_JOB_BUILDCACHE_NAME=`spack -d buildcache get-buildcache-name --spec-yaml "${DEP_SPEC_YAML_PATH}"`
if [[ $? -eq 0 ]]; then
DEP_JOB_ID_FILE="${BUILD_CACHE_DIR}/${DEP_JOB_BUILDCACHE_NAME}.cdashid"
echo "DEP_JOB_ID_FILE path = ${DEP_JOB_ID_FILE}"
if [ -f "${DEP_JOB_ID_FILE}" ]; then
DEP_JOB_CDASH_BUILD_ID=$(<${DEP_JOB_ID_FILE})
echo "File ${DEP_JOB_ID_FILE} contained value ${DEP_JOB_CDASH_BUILD_ID}"
echo "Relating builds -> ${JOB_SPEC_NAME} (buildid=${JOB_CDASH_BUILD_ID}) depends on ${DEP_PKG_NAME} (buildid=${DEP_JOB_CDASH_BUILD_ID})"
relateBuildsPostBody="$(get_relate_builds_post_data "Spack" ${JOB_CDASH_BUILD_ID} ${DEP_JOB_CDASH_BUILD_ID})"
relateBuildsResult=`curl "${DEP_JOB_RELATEBUILDS_URL}" -H "Content-Type: application/json" -H "Accept: application/json" -d "${relateBuildsPostBody}"`
echo "Result of curl request: ${relateBuildsResult}"
else
echo "ERROR: Did not find expected .cdashid file for dependency: ${DEP_JOB_ID_FILE}"
exit 1
fi
else
echo "ERROR: Unable to get buildcache entry name for ${DEP_SPEC_NAME}"
exit 1
fi
done
else
echo "ERROR: Did not find expected .cdashid file ${JOB_CDASH_ID_FILE}"
exit 1
fi
# Show the size of the buildcache and a list of what's in it, directly
# in the gitlab log output
(
restore_io
du -sh ${BUILD_CACHE_DIR}
find ${BUILD_CACHE_DIR} -maxdepth 3 -type d -ls
)
echo "End of rebuild package script"

View file

@ -0,0 +1,212 @@
# Copyright 2013-2019 Lawrence Livermore National Security, LLC and other
# Spack Project Developers. See the top-level COPYRIGHT file for details.
#
# SPDX-License-Identifier: (Apache-2.0 OR MIT)
# TODO: This will be merged into the buildcache command once
# everything is working.
import os
import re
import sys
try:
import boto3
import botocore
have_boto3_support = True
except ImportError:
have_boto3_support = False
import llnl.util.tty as tty
from spack.error import SpackError
import spack.tengine as template_engine
from spack.spec import Spec
import spack.binary_distribution as bindist
description = "temporary command to upload buildcaches to 's3.spack.io'"
section = "packaging"
level = "long"
def setup_parser(subparser):
setup_parser.parser = subparser
subparsers = subparser.add_subparsers(help='upload-s3 sub-commands')
# sub-command to upload a built spec to s3
spec = subparsers.add_parser('spec', help=upload_spec.__doc__)
spec.add_argument('-s', '--spec', default=None,
help='Spec to upload')
spec.add_argument('-y', '--spec-yaml', default=None,
help='Path to spec yaml file containing spec to upload')
spec.add_argument('-b', '--base-dir', default=None,
help='Path to root of buildcaches')
spec.add_argument('-e', '--endpoint-url',
default='https://s3.spack.io', help='URL of mirror')
spec.set_defaults(func=upload_spec)
# sub-command to update the index of a buildcache on s3
index = subparsers.add_parser('index', help=update_index.__doc__)
index.add_argument('-e', '--endpoint-url',
default='https://s3.spack.io', help='URL of mirror')
index.set_defaults(func=update_index)
def get_s3_session(endpoint_url):
if not have_boto3_support:
raise SpackError('boto3 module not available')
session = boto3.Session()
s3 = session.resource('s3')
bucket_names = []
for bucket in s3.buckets.all():
bucket_names.append(bucket.name)
if len(bucket_names) > 1:
raise SpackError('More than one bucket associated with credentials')
bucket_name = bucket_names[0]
return s3, bucket_name
def update_index(args):
"""Update the index of an s3 buildcache"""
s3, bucket_name = get_s3_session(args.endpoint_url)
bucket = s3.Bucket(bucket_name)
exists = True
try:
s3.meta.client.head_bucket(Bucket=bucket_name)
except botocore.exceptions.ClientError as e:
# If a client error is thrown, then check that it was a 404 error.
# If it was a 404 error, then the bucket does not exist.
error_code = e.response['Error']['Code']
if error_code == '404':
exists = False
if not exists:
tty.error('S3 bucket "{0}" does not exist'.format(bucket_name))
sys.exit(1)
build_cache_dir = os.path.join(
'mirror', bindist.build_cache_relative_path())
spec_yaml_regex = re.compile('{0}/(.+\\.spec\\.yaml)$'.format(
build_cache_dir))
spack_regex = re.compile('{0}/([^/]+)/.+\\.spack$'.format(
build_cache_dir))
top_level_keys = set()
for key in bucket.objects.all():
m = spec_yaml_regex.search(key.key)
if m:
top_level_keys.add(m.group(1))
print(m.group(1))
continue
m = spack_regex.search(key.key)
if m:
top_level_keys.add(m.group(1))
print(m.group(1))
continue
index_data = {
'top_level_keys': top_level_keys,
}
env = template_engine.make_environment()
template_dir = 'misc'
index_template = os.path.join(template_dir, 'buildcache_index.html')
t = env.get_template(index_template)
contents = t.render(index_data)
index_key = os.path.join(build_cache_dir, 'index.html')
tty.debug('Generated index:')
tty.debug(contents)
tty.debug('Pushing it to {0} -> {1}'.format(bucket_name, index_key))
s3_obj = s3.Object(bucket_name, index_key)
s3_obj.put(Body=contents, ACL='public-read')
def upload_spec(args):
"""Upload a spec to s3 bucket"""
if not args.spec and not args.spec_yaml:
tty.error('Cannot upload spec without spec arg or path to spec yaml')
sys.exit(1)
if not args.base_dir:
tty.error('No base directory for buildcache specified')
sys.exit(1)
if args.spec:
try:
spec = Spec(args.spec)
spec.concretize()
except Exception:
tty.error('Unable to concrectize spec from string {0}'.format(
args.spec))
sys.exit(1)
else:
try:
with open(args.spec_yaml, 'r') as fd:
spec = Spec.from_yaml(fd.read())
except Exception:
tty.error('Unable to concrectize spec from yaml {0}'.format(
args.spec_yaml))
sys.exit(1)
s3, bucket_name = get_s3_session(args.endpoint_url)
build_cache_dir = bindist.build_cache_relative_path()
tarball_key = os.path.join(
build_cache_dir, bindist.tarball_path_name(spec, '.spack'))
tarball_path = os.path.join(args.base_dir, tarball_key)
specfile_key = os.path.join(
build_cache_dir, bindist.tarball_name(spec, '.spec.yaml'))
specfile_path = os.path.join(args.base_dir, specfile_key)
cdashidfile_key = os.path.join(
build_cache_dir, bindist.tarball_name(spec, '.cdashid'))
cdashidfile_path = os.path.join(args.base_dir, cdashidfile_key)
tty.msg('Uploading {0}'.format(tarball_key))
s3.meta.client.upload_file(
tarball_path, bucket_name,
os.path.join('mirror', tarball_key),
ExtraArgs={'ACL': 'public-read'})
tty.msg('Uploading {0}'.format(specfile_key))
s3.meta.client.upload_file(
specfile_path, bucket_name,
os.path.join('mirror', specfile_key),
ExtraArgs={'ACL': 'public-read'})
if os.path.exists(cdashidfile_path):
tty.msg('Uploading {0}'.format(cdashidfile_key))
s3.meta.client.upload_file(
cdashidfile_path, bucket_name,
os.path.join('mirror', cdashidfile_key),
ExtraArgs={'ACL': 'public-read'})
def upload_s3(parser, args):
if args.func:
args.func(args)