#!/bin/bash # Copyright 2015 The Kubernetes Authors. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. # This script uploads metadata and test results to Google Cloud Storage, in the # location indicated by JENKINS_GCS_LOGS_PATH. By default, we use the Google # kubernetes-jenkins bucket. # # The script looks for one of two environment variables to be set: # JENKINS_BUILD_STARTED: set to a nonempty string to upload version # information to 'started.json'. The value of the variable is not # currently used. # JENKINS_BUILD_FINISHED: set to the Jenkins build result to upload the build # result to 'finished.json', any test artifacts, and update the # 'latest-build.txt' file pointer. Since this script uses gsutil directly, # it's a bit faster at uploading large numbers of files than the GCS Jenkins # plugin. It also makes use of gsutil's gzip functionality. # # Note: for magicfile support to work correctly, the "file" utility must be # installed. # TODO(rmmh): rewrite this script in Python so we can actually test it! set -o errexit set -o nounset set -o pipefail if [[ -n "${JENKINS_BUILD_STARTED:-}" && -n "${JENKINS_BUILD_FINISHED:-}" ]]; then echo "Error: JENKINS_BUILD_STARTED and JENKINS_BUILD_FINISHED should not both be set!" exit 1 fi if [[ ! ${JENKINS_UPLOAD_TO_GCS:-y} =~ ^[yY]$ ]]; then exit 0 fi # Attempt to determine if we're running against a repo other than # kubernetes/kubernetes to determine whether to place PR logs in a different # location. # # In the current CI system, the tracked repo is named remote. This is not true # in general for most devs, where origin and upstream are more common. GCS_SUBDIR="" readonly remote_git_repo=$(git config --get remote.remote.url | sed 's:.*github.com/::' || true) if [[ -n "${remote_git_repo}" ]]; then case "${remote_git_repo}" in # main repo: nothing extra kubernetes/kubernetes) GCS_SUBDIR="" ;; # a different repo on the k8s org: just the repo name (strip kubernetes/) kubernetes/*) GCS_SUBDIR="${remote_git_repo#kubernetes/}/" ;; # any other repo: ${org}_${repo} (replace / with _) *) GCS_SUBDIR="${remote_git_repo/\//_}/" ;; esac if [[ "${remote_git_repo}" != "kubernetes/kubernetes" ]]; then # also store the repo in started.json, so Gubernator can link it properly. export BUILD_METADATA_REPO="${remote_git_repo}" fi fi if [[ ${JOB_NAME} =~ -pull- ]]; then : ${JENKINS_GCS_LOGS_PATH:="gs://kubernetes-jenkins/pr-logs/pull/${GCS_SUBDIR}${ghprbPullId:-unknown}"} : ${JENKINS_GCS_LATEST_PATH:="gs://kubernetes-jenkins/pr-logs/directory"} : ${JENKINS_GCS_LOGS_INDIRECT:="gs://kubernetes-jenkins/pr-logs/directory/${JOB_NAME}"} else : ${JENKINS_GCS_LOGS_PATH:="gs://kubernetes-jenkins/logs"} : ${JENKINS_GCS_LATEST_PATH:="gs://kubernetes-jenkins/logs"} : ${JENKINS_GCS_LOGS_INDIRECT:=""} fi readonly artifacts_path="${WORKSPACE}/_artifacts" readonly gcs_job_path="${JENKINS_GCS_LOGS_PATH}/${JOB_NAME}" readonly gcs_build_path="${gcs_job_path}/${BUILD_NUMBER}" readonly gcs_latest_path="${JENKINS_GCS_LATEST_PATH}/${JOB_NAME}" readonly gcs_indirect_path="${JENKINS_GCS_LOGS_INDIRECT}" readonly gcs_acl="public-read" readonly results_url=${gcs_build_path//"gs:/"/"https://console.cloud.google.com/storage/browser"} readonly timestamp=$(date +%s) ######################################################################### # $0 is called from different contexts so figure out where kubernetes is. # Sets non-exported global kubernetes_base_path and defaults to "." function set_kubernetes_base_path () { for kubernetes_base_path in kubernetes go/src/k8s.io/kubernetes .; do # Pick a canonical item to find in a kubernetes tree which could be a # raw source tree or an expanded tarball. [[ -f ${kubernetes_base_path}/cluster/common.sh ]] && break done } ######################################################################### # Try to discover the kubernetes version. # prints version function find_version() { ( # Where are we? # This could be set in the global scope at some point if we need to # discover the kubernetes path elsewhere. set_kubernetes_base_path cd ${kubernetes_base_path} if [[ -e "version" ]]; then cat version elif [[ -e "hack/lib/version.sh" ]]; then export KUBE_ROOT="." source "hack/lib/version.sh" kube::version::get_version_vars echo "${KUBE_GIT_VERSION-}" else # Last resort from the started.json gsutil cat ${gcs_build_path}/started.json 2>/dev/null |\ sed -n 's/ *"version": *"\([^"]*\)",*/\1/p' fi ) } # Output started.json. Use test function below! function print_started() { local metadata_keys=$(compgen -e | grep ^BUILD_METADATA_) echo "{" echo " \"version\": \"${version}\"," echo " \"timestamp\": ${timestamp}," if [[ -n "${metadata_keys}" ]]; then # Any exported variables of the form BUILD_METADATA_KEY=VALUE # will be available as started["metadata"][KEY.lower()]. echo " \"metadata\": {" local sep="" # leading commas are easy to track for env_var in $metadata_keys; do local var_upper="${env_var#BUILD_METADATA_}" echo " $sep\"${var_upper,,}\": \"${!env_var}\"" sep="," done echo " }," fi echo " \"jenkins-node\": \"${NODE_NAME:-}\"" echo "}" } # Use this to test changes to print_started. if [[ -n "${TEST_STARTED_JSON:-}" ]]; then version=$(find_version) cat <(print_started) | jq . exit fi function upload_version() { local -r version=$(find_version) local upload_attempt echo -n 'Run starting at '; date -d "@${timestamp}" if [[ -n "${version}" ]]; then echo "Found Kubernetes version: ${version}" else echo "Could not find Kubernetes version" fi local -r json_file="${gcs_build_path}/started.json" for upload_attempt in {1..3}; do echo "Uploading version to: ${json_file} (attempt ${upload_attempt})" gsutil -q -h "Content-Type:application/json" cp -a "${gcs_acl}" <(print_started) "${json_file}" || continue break done } ######################################################################### # Maintain a single file storing the full build version, Jenkins' job number # build state. Limit its size so it does not grow unbounded. # This is primarily used for and by the # github.com/kubernetes/release/find_green_build tool. # @param build_result - the state of the build # function update_job_result_cache() { local -r build_result=$1 local -r version=$(find_version) local -r job_results=${gcs_job_path}/jobResultsCache.json local -r tmp_results="${WORKSPACE}/_tmp/jobResultsCache.tmp" # TODO: This constraint is insufficient. The boundary for secondary # job cache should be date based on the last primary build. # The issue is we are trying to find a matched green set of results # at a given hash, but all of the jobs run at wildly different lengths. local -r cache_size=300 local upload_attempt if [[ -n "${version}" ]]; then echo "Found Kubernetes version: ${version}" else echo "Could not find Kubernetes version" fi mkdir -p ${tmp_results%/*} # Construct a valid json file echo "[" > ${tmp_results} for upload_attempt in $(seq 3); do echo "Copying ${job_results} to ${tmp_results} (attempt ${upload_attempt})" # The sed construct below is stripping out only the "version" lines # and then ensuring there's a single comma at the end of the line. gsutil -q cat ${job_results} 2>&- |\ sed -n 's/^\({"version".*}\),*/\1,/p' |\ tail -${cache_size} >> ${tmp_results} || continue break done echo "{\"version\": \"${version}\", \"buildnumber\": \"${BUILD_NUMBER}\"," \ "\"result\": \"${build_result}\"}" >> ${tmp_results} echo "]" >> ${tmp_results} for upload_attempt in $(seq 3); do echo "Copying ${tmp_results} to ${job_results} (attempt ${upload_attempt})" gsutil -q -h "Content-Type:application/json" cp -a "${gcs_acl}" \ ${tmp_results} ${job_results} || continue break done rm -f ${tmp_results} } function upload_artifacts_and_build_result() { local -r build_result=$1 local upload_attempt echo -n 'Run finished at '; date -d "@${timestamp}" for upload_attempt in {1..3}; do echo "Uploading to ${gcs_build_path} (attempt ${upload_attempt})" echo "Uploading build result: ${build_result}" gsutil -q -h "Content-Type:application/json" cp -a "${gcs_acl}" <( echo "{" echo " \"result\": \"${build_result}\"," echo " \"timestamp\": ${timestamp}" echo "}" ) "${gcs_build_path}/finished.json" || continue if [[ -d "${artifacts_path}" && -n $(ls -A "${artifacts_path}") ]]; then echo "Uploading artifacts" gsutil -m -q -o "GSUtil:use_magicfile=True" cp -a "${gcs_acl}" -r -c \ -z log,txt,xml "${artifacts_path}" "${gcs_build_path}/artifacts" || continue fi if [[ -e "${WORKSPACE}/build-log.txt" ]]; then echo "Uploading build log" gsutil -q cp -Z -a "${gcs_acl}" "${WORKSPACE}/build-log.txt" "${gcs_build_path}" fi # For pull jobs, keep a canonical ordering for tools that want to examine # the output. if [[ "${gcs_indirect_path}" != "" ]]; then echo "Writing ${gcs_build_path} to ${gcs_indirect_path}/${BUILD_NUMBER}.txt" echo "${gcs_build_path}" | \ gsutil -q -h "Content-Type:text/plain" \ cp -a "${gcs_acl}" - "${gcs_indirect_path}/${BUILD_NUMBER}.txt" || continue echo "Marking build ${BUILD_NUMBER} as the latest completed build for this PR" echo "${BUILD_NUMBER}" | \ gsutil -q -h "Content-Type:text/plain" -h "Cache-Control:private, max-age=0, no-transform" \ cp -a "${gcs_acl}" - "${gcs_job_path}/latest-build.txt" || continue fi # Mark this build as the latest completed. echo "Marking build ${BUILD_NUMBER} as the latest completed build" echo "${BUILD_NUMBER}" | \ gsutil -q -h "Content-Type:text/plain" -h "Cache-Control:private, max-age=0, no-transform" \ cp -a "${gcs_acl}" - "${gcs_latest_path}/latest-build.txt" || continue break # all uploads succeeded if we hit this point done echo -e "\n\n\n*** View logs and artifacts at ${results_url} ***\n\n" } if [[ -n "${JENKINS_BUILD_STARTED:-}" ]]; then upload_version elif [[ -n "${JENKINS_BUILD_FINISHED:-}" ]]; then upload_artifacts_and_build_result ${JENKINS_BUILD_FINISHED} update_job_result_cache ${JENKINS_BUILD_FINISHED} else echo "Called without JENKINS_BUILD_STARTED or JENKINS_BUILD_FINISHED set." echo "Assuming a legacy invocation." upload_artifacts_and_build_result "[UNSET]" fi