sonic-buildimage/azure-pipelines.yml
Yutong Zhang cb354a5af2
[TestbedV2][master] Remove timeout in each step. (#12915)
Previously, we set timeout in each step such as Lock testbed, Prepare testbed, Run test and KVM dump. When some issue suck like retry happens in one step, it will cause timeout error, but actually, it only needs more time to success. In this pr, we remove the timeout limit in each step and control the timeout outside in each job. When the job runs more than four hours, it will be cancelled.

Why I did it
Previously, we set timeout in each step such as Lock testbed, Prepare testbed, Run test and KVM dump. When some issue suck like retry happens in one step, it will cause timeout error, but actually, it only needs more time to success. In this pr, we remove the timeout limit in each step and control the timeout outside in each job. When the job runs more than four hours, it will be cancelled.

How I did it
Remove the timeout parameter in each step, and control the timeout outside in each job.

How to verify it
Set the timeout of one job to 4 hours, and when timeout happens, azure pipeline will cancel this job.
2022-12-03 22:30:03 -08:00

383 lines
12 KiB
YAML

# Starter pipeline
# Start with a minimal pipeline that you can customize to build and deploy your code.
# Add steps that build, run tests, deploy, and more:
# https://aka.ms/yaml
trigger:
branches:
include:
- master
- 202???
paths:
exclude:
- .github
pr:
branches:
include:
- master
- 202???
paths:
exclude:
- .github
name: $(TeamProject)_$(Build.DefinitionName)_$(SourceBranchName)_$(Date:yyyyMMdd)$(Rev:.r)
resources:
repositories:
- repository: sonic-mgmt
type: github
name: sonic-net/sonic-mgmt
endpoint: sonic-net
- repository: buildimage
type: github
name: sonic-net/sonic-buildimage
endpoint: sonic-net
ref: master
variables:
- template: .azure-pipelines/azure-pipelines-repd-build-variables.yml@buildimage
- template: .azure-pipelines/template-variables.yml@buildimage
- name: CACHE_MODE
value: rcache
- name: ENABLE_FIPS
value: y
stages:
- stage: BuildVS
pool: sonicbld
jobs:
- template: .azure-pipelines/azure-pipelines-build.yml
parameters:
buildOptions: 'USERNAME=admin SONIC_BUILD_JOBS=$(nproc) BUILD_MULTIASIC_KVM=y ${{ variables.VERSION_CONTROL_OPTIONS }}'
jobGroups:
- name: vs
- stage: Build
pool: sonicbld
dependsOn: []
jobs:
- template: .azure-pipelines/azure-pipelines-build.yml
parameters:
buildOptions: 'USERNAME=admin SONIC_BUILD_JOBS=$(nproc) ${{ variables.VERSION_CONTROL_OPTIONS }}'
jobGroups:
- name: broadcom
variables:
swi_image: yes
- name: mellanox
- name: marvell-armhf
pool: sonicbld-armhf
timeoutInMinutes: 1200
variables:
PLATFORM_ARCH: armhf
- stage: Test
dependsOn: BuildVS
condition: and(succeeded(), and(ne(stageDependencies.BuildVS.outputs['vs.SetVar.SKIP_VSTEST'], 'YES'), in(dependencies.BuildVS.result, 'Succeeded', 'SucceededWithIssues')))
variables:
- group: Testbed-Tools
- name: inventory
value: veos_vtb
- name: testbed_file
value: vtestbed.csv
# For every test job:
# continueOnError: false means it's a required test job and will block merge if it fails
# continueOnError: true means it's an optional test job and will not block merge even though it fails(unless a required test job depends on its result)
jobs:
- job:
pool: sonictest
displayName: "vstest"
timeoutInMinutes: 60
steps:
- template: .azure-pipelines/cleanup.yml
- checkout: self
clean: true
submodules: recursive
displayName: 'Checkout code'
- script: |
sudo rm -rf ../target
sudo rm -rf ../*.deb
displayName: "Cleanup"
- task: DownloadPipelineArtifact@2
inputs:
source: specific
project: build
pipeline: 9
artifact: sonic-swss-common.amd64.ubuntu20_04
runVersion: 'latestFromBranch'
runBranch: 'refs/heads/master'
displayName: "Download sonic swss common deb packages"
- task: DownloadPipelineArtifact@2
inputs:
artifact: sonic-buildimage.vs
displayName: "Download sonic-buildimage.vs artifact"
- script: |
set -x
sudo apt-get update
sudo apt-get install libyang0.16 -y
sudo dpkg -i --force-confask,confnew ../libswsscommon_1.0.0_amd64.deb
sudo dpkg -i ../python3-swsscommon_1.0.0_amd64.deb
sudo docker load -i ../target/docker-sonic-vs.gz
docker tag docker-sonic-vs:latest docker-sonic-vs:$(Build.BuildNumber)
username=$(id -un)
trap "docker ps; docker images; ip netns list; \
docker rmi docker-sonic-vs:$(Build.BuildNumber); \
ip netns list | grep -E [-]srv[0-9]+ | awk '{print $1}' | xargs -I {} sudo ip netns delete {}; \
sudo chown -R ${username}.${username} .; \
sudo chown -R ${username}.${username} $(System.DefaultWorkingDirectory)" EXIT
pushd platform/vs/tests
sudo py.test -v --junitxml=tr.xml --imgname=docker-sonic-vs:$(Build.BuildNumber)
displayName: "Run vs tests"
- task: PublishTestResults@2
inputs:
testResultsFiles: '**/tr.xml'
testRunTitle: vstest
- job: t0_part1
pool: sonictest
displayName: "kvmtest-t0-part1"
timeoutInMinutes: 400
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_CLASSICAL_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-template.yml
parameters:
dut: vlab-01
tbname: vms-kvm-t0
ptf_name: ptf_vms6-1
tbtype: t0
vmtype: ceos
section: part-1
- job: t0_part2
pool: sonictest
displayName: "kvmtest-t0-part2"
timeoutInMinutes: 400
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_CLASSICAL_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-template.yml
parameters:
dut: vlab-01
tbname: vms-kvm-t0
ptf_name: ptf_vms6-1
tbtype: t0
vmtype: ceos
section: part-2
- job: t0_testbedv2
pool:
vmImage: 'ubuntu-20.04'
displayName: "kvmtest-t0 by TestbedV2"
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: t0
MIN_WORKER: $(T0_INSTANCE_NUM)
MAX_WORKER: $(T0_INSTANCE_NUM)
- job: t0_2vlans_testbedv2
pool:
vmImage: 'ubuntu-20.04'
displayName: "kvmtest-t0-2vlans by TestbedV2"
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: t0
TEST_SET: t0-2vlans
MIN_WORKER: $(T0_2VLANS_INSTANCE_NUM)
MAX_WORKER: $(T0_2VLANS_INSTANCE_NUM)
DEPLOY_MG_EXTRA_PARAMS: "-e vlan_config=two_vlan_a"
- job:
pool:
vmImage: 'ubuntu-20.04'
displayName: "kvmtest-t0"
dependsOn:
- t0_part1
- t0_part2
- t0_testbedv2
- t0_2vlans_testbedv2
condition: always()
continueOnError: false
variables:
resultOfPart1: $[ dependencies.t0_part1.result ]
resultOfPart2: $[ dependencies.t0_part2.result ]
resultOfT0TestbedV2: $[ dependencies.t0_testbedv2.result ]
resultOfT02VlansTestbedV2: $[ dependencies.t0_2vlans_testbedv2.result ]
steps:
- script: |
if [ $(resultOfT0TestbedV2) == "Succeeded" ] && [ $(resultOfT02VlansTestbedV2) == "Succeeded" ]; then
echo "TestbedV2 t0 passed."
exit 0
fi
if [ $(resultOfPart1) == "Succeeded" ] && [ $(resultOfPart2) == "Succeeded" ]; then
echo "Classic t0 jobs(both part1 and part2) passed."
exit 0
fi
echo "Both classic and TestbedV2 t0 jobs failed! Please check the detailed information. (Any of them passed, t0 will be considered as passed)"
exit 1
- job: t1_lag_classic
pool: sonictest-t1-lag
displayName: "kvmtest-t1-lag classic"
timeoutInMinutes: 400
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_CLASSICAL_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-template.yml
parameters:
dut: vlab-03
tbname: vms-kvm-t1-lag
ptf_name: ptf_vms6-2
tbtype: t1-lag
vmtype: ceos
- job: t1_lag_testbedv2
pool:
vmImage: 'ubuntu-20.04'
displayName: "kvmtest-t1-lag by TestbedV2"
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: t1-lag
MIN_WORKER: $(T1_LAG_INSTANCE_NUM)
MAX_WORKER: $(T1_LAG_INSTANCE_NUM)
- job:
pool:
vmImage: 'ubuntu-20.04'
displayName: "kvmtest-t1-lag"
dependsOn:
- t1_lag_classic
- t1_lag_testbedv2
condition: always()
continueOnError: false
variables:
resultOfClassic: $[ dependencies.t1_lag_classic.result ]
resultOfTestbedV2: $[ dependencies.t1_lag_testbedv2.result ]
steps:
- script: |
if [ $(resultOfClassic) == "Succeeded" ] || [ $(resultOfTestbedV2) == "Succeeded" ]; then
echo "One or both of t1_lag_classic and t1_lag_testbedv2 passed."
exit 0
else
echo "Both t1_lag_classic and t1_lag_testbedv2 failed! Please check the detailed information."
exit 1
fi
- job:
pool: sonictest-sonic-t0
displayName: "kvmtest-t0-sonic"
timeoutInMinutes: 360
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_CLASSICAL_TEST, 'YES'))
continueOnError: true
steps:
- template: .azure-pipelines/run-test-template.yml
parameters:
dut: vlab-02
tbname: vms-kvm-t0-64-32
ptf_name: ptf_vms6-1
tbtype: t0-sonic
vmtype: vsonic
- job:
pool: sonictest-ma
displayName: "kvmtest-multi-asic-t1-lag"
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_CLASSICAL_TEST, 'YES'))
continueOnError: true
steps:
- template: .azure-pipelines/run-test-template.yml
parameters:
dut: vlab-08
tbname: vms-kvm-four-asic-t1-lag
ptf_name: ptf_vms6-4
tbtype: multi-asic-t1-lag-pr
image: sonic-4asic-vs.img.gz
- job: multi_asic_testbedv2
displayName: "kvmtest-multi-asic-t1-lag by TestbedV2"
pool:
vmImage: 'ubuntu-20.04'
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: true
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: t1-8-lag
TEST_SET: multi-asic-t1-lag
MIN_WORKER: $(MULTI_ASIC_INSTANCE_NUM)
MAX_WORKER: $(MULTI_ASIC_INSTANCE_NUM)
NUM_ASIC: 4
- job: dualtor_testbedv2
pool:
vmImage: 'ubuntu-20.04'
displayName: "kvmtest-dualtor-t0 by TestbedV2"
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: false
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: dualtor
MIN_WORKER: $(T0_DUALTOR_INSTANCE_NUM)
MAX_WORKER: $(T0_DUALTOR_INSTANCE_NUM)
COMMON_EXTRA_PARAMS: "--disable_loganalyzer "
- job: sonic_t0_testbedv2
displayName: "kvmtest-t0-sonic by TestbedV2"
pool:
vmImage: 'ubuntu-20.04'
timeoutInMinutes: 240
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: true
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: t0-64-32
MIN_WORKER: $(T0_SONIC_INSTANCE_NUM)
MAX_WORKER: $(T0_SONIC_INSTANCE_NUM)
TEST_SET: t0-sonic
COMMON_EXTRA_PARAMS: "--neighbor_type=sonic --enable_macsec --macsec_profile=128_SCI,256_XPN_SCI"
VM_TYPE: vsonic
SPECIFIED_PARAMS: '{\"test_pretest.py\":[\"--completeness_level=confident\",\"--allow_recover\"],\"test_posttest.py\":[\"--completeness_level=confident\",\"--allow_recover\"]}'
- job: wan_testbedv2
displayName: "kvmtest-wan by TestbedV2"
pool:
vmImage: 'ubuntu-20.04'
timeoutInMinutes: 1080
condition: and(succeeded(), eq(variables.BUILD_IMG_RUN_TESTBEDV2_TEST, 'YES'))
continueOnError: true
steps:
- template: .azure-pipelines/run-test-scheduler-template.yml
parameters:
TOPOLOGY: wan-pub
MIN_WORKER: $(WAN_INSTANCE_NUM)
MAX_WORKER: $(WAN_INSTANCE_NUM)
COMMON_EXTRA_PARAMS: "--skip_sanity "