This is an automated email from the ASF dual-hosted git repository.
justinchen pushed a commit to branch fix-login
in repository https://gitbox.apache.org/repos/asf/iotdb.git
The following commit(s) were added to refs/heads/fix-login by this push:
new 4b275c1ff94 Update pipe-it.yml
4b275c1ff94 is described below
commit 4b275c1ff940c0f806f3f58a0916684c16bb7c22
Author: Caideyipi <[email protected]>
AuthorDate: Mon Mar 23 10:37:40 2026 +0800
Update pipe-it.yml
---
.github/workflows/pipe-it.yml | 988 +++---------------------------------------
1 file changed, 58 insertions(+), 930 deletions(-)
diff --git a/.github/workflows/pipe-it.yml b/.github/workflows/pipe-it.yml
index 0968e7739a0..1c1958ad1dc 100644
--- a/.github/workflows/pipe-it.yml
+++ b/.github/workflows/pipe-it.yml
@@ -1,31 +1,16 @@
-name: Multi-Cluster IT
+name: Pipe IT Debug (Single Test Loop)
on:
push:
branches:
- - master
- - "rel/*"
- - "rc/*"
- paths-ignore:
- - "docs/**"
- - "site/**"
- - "iotdb-core/datanode/src/main/java/org/apache/iotdb/db/queryengine/**"
#queryengine
+ - "**" # 只要有代码推送就跑，方便你在 PR 上触发
pull_request:
branches:
- - master
- - "rel/*"
- - "rc/*"
- - "force_ci/**"
- paths-ignore:
- - "docs/**"
- - "site/**"
- - "iotdb-core/datanode/src/main/java/org/apache/iotdb/db/queryengine/**"
#queryengine
- # allow manually run the action:
- workflow_dispatch:
+ - "**"
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
- cancel-in-progress: true
+ cancel-in-progress: true # 如果有新的提交，取消旧的构建，节省资源
env:
MAVEN_OPTS: -Dhttp.keepAlive=false -Dmaven.wagon.http.pool=false
-Dmaven.wagon.http.retryHandler.class=standard
-Dmaven.wagon.http.retryHandler.count=3
@@ -33,104 +18,18 @@ env:
DEVELOCITY_ACCESS_KEY: ${{ secrets.DEVELOCITY_ACCESS_KEY }}
jobs:
- single:
+ debug-specific-it:
+ name: Run IoTDBPipePermissionIT in Loop
+ runs-on: ubuntu-latest
strategy:
fail-fast: false
- max-parallel: 15
matrix:
java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
cluster1: [HighPerformanceMode]
cluster2: [HighPerformanceMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
steps:
- uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
- -pl integration-test \
- -am -PMultiClusterIT1 \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-single-java${{ matrix.java }}-${{ runner.os }}-${{
matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- dual-tree-auto-basic:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster: [HighPerformanceMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- name: Set up JDK ${{ matrix.java }}
uses: actions/setup-java@v5
with:
@@ -138,841 +37,70 @@ jobs:
java-version: ${{ matrix.java }}
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster }},${{ matrix.cluster
}} \
- -pl integration-test \
- -am -PMultiClusterIT2DualTreeAutoBasic \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-dual-tree-auto-basic-java${{ matrix.java }}-${{
runner.os }}-${{ matrix.cluster }}-${{ matrix.cluster }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- dual-tree-auto-enhanced:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster1: [HighPerformanceMode]
- cluster2: [HighPerformanceMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- name: Cache Maven packages
uses: actions/cache@v5
with:
path: ~/.m2
key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
- -pl integration-test \
- -am -PMultiClusterIT2DualTreeAutoEnhanced \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-dual-tree-auto-enhanced-java${{ matrix.java }}-${{
runner.os }}-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- dual-tree-manual:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster1: [HighPerformanceMode]
- cluster2: [HighPerformanceMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
- -pl integration-test \
- -am -PMultiClusterIT2DualTreeManual \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-dual-tree-manual-java${{ matrix.java }}-${{
runner.os }}-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- subscription-tree-arch-verification:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster1: [ScalableSingleNodeMode]
- cluster2: [ScalableSingleNodeMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
+ - name: Run Test in Loop Until Failure
shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
+ # 确保日志目录存在
+ mkdir -p integration-test/target/cluster-logs
+
+ local -i attempt=1
+
+ while true; do
+ echo "===================================================="
+ echo "Starting attempt #$attempt"
+ echo "===================================================="
+
+ # 注意：这里使用 mvn verify 或 test，重点是 -Dtest 参数
+ # 请根据你的项目模块结构调整 -pl 参数，如果 integration-test 是根模块则不需要 -pl
+ # 我保留了你原有的 -PMultiClusterIT2DualTreeManual Profile，请确认这是正确的 Profile
+
+ mvn clean verify \
+ -P with-integration-tests,MultiClusterIT2DualTreeManual \
-DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
+ -Dtest=IoTDBPipePermissionIT#testIllegalPassword \
+ -DintegrationTest.forkCount=1 \
+ -DConfigNodeMaxHeapSize=256 \
+ -DDataNodeMaxHeapSize=1024 \
+ -DDataNodeMaxDirectMemorySize=768 \
-DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
-pl integration-test \
- -am -PMultiClusterIT2SubscriptionTreeArchVerification \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-subscription-tree-arch-verification-java${{
matrix.java }}-${{ runner.os }}-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- subscription-table-arch-verification:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster1: [ScalableSingleNodeMode]
- cluster2: [ScalableSingleNodeMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
- -pl integration-test \
- -am -PMultiClusterIT2SubscriptionTableArchVerification \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-subscription-table-arch-verification-java${{
matrix.java }}-${{ runner.os }}-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- subscription-tree-regression-consumer:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # do not use HighPerformanceMode here, otherwise some tests will cause
the GH runner to receive a shutdown signal
- cluster1: [ScalableSingleNodeMode]
- cluster2: [ScalableSingleNodeMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
- -pl integration-test \
- -am -PMultiClusterIT2SubscriptionTreeRegressionConsumer \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-subscription-tree-regression-consumer-java${{
matrix.java }}-${{ runner.os }}-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- subscription-tree-regression-misc:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # do not use HighPerformanceMode here, otherwise some tests will cause
the GH runner to receive a shutdown signal
- cluster1: [ScalableSingleNodeMode]
- cluster2: [ScalableSingleNodeMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }} \
- -pl integration-test \
- -am -PMultiClusterIT2SubscriptionTreeRegressionMisc \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-subscription-tree-regression-misc-java${{
matrix.java }}-${{ runner.os }}-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- dual-table-manual-basic:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster: [HighPerformanceMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster }},${{ matrix.cluster
}} \
- -pl integration-test \
- -am -PMultiClusterIT2DualTableManualBasic \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-dual-table-manual-basic-java${{ matrix.java }}-${{
runner.os }}-${{ matrix.cluster }}-${{ matrix.cluster }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- dual-table-manual-enhanced:
- strategy:
- fail-fast: false
- max-parallel: 15
- matrix:
- java: [17]
- # StrongConsistencyClusterMode is ignored now because RatisConsensus
has not been supported yet.
- cluster: [HighPerformanceMode]
- os: [ubuntu-latest]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster }},${{ matrix.cluster
}} \
- -pl integration-test \
- -am -PMultiClusterIT2DualTableManualEnhanced \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
- if: failure()
- uses: actions/upload-artifact@v6
- with:
- name: cluster-log-dual-table-manual-enhanced-java${{ matrix.java
}}-${{ runner.os }}-${{ matrix.cluster }}-${{ matrix.cluster }}
- path: integration-test/target/cluster-logs
- retention-days: 30
- triple:
- strategy:
- fail-fast: false
- max-parallel: 1
- matrix:
- java: [ 17 ]
- cluster1: [ ScalableSingleNodeMode ]
- cluster2: [ ScalableSingleNodeMode ]
- cluster3: [ ScalableSingleNodeMode ]
- os: [ ubuntu-latest ]
- runs-on: ${{ matrix.os }}
- steps:
- - uses: actions/checkout@v5
- - name: Set up JDK ${{ matrix.java }}
- uses: actions/setup-java@v5
- with:
- distribution: corretto
- java-version: ${{ matrix.java }}
- env:
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
- - name: Cache Maven packages
- uses: actions/cache@v5
- with:
- path: ~/.m2
- key: ${{ runner.os }}-m2-${{ hashFiles('**/pom.xml') }}
- restore-keys: ${{ runner.os }}-m2-
- - name: Sleep for a random duration between 0 and 10000 milliseconds
- run: |
- sleep $(( $(( RANDOM % 10000 + 1 )) / 1000))
- - name: IT Test
- shell: bash
- # we do not compile client-cpp for saving time, it is tested in
client.yml
- # we can skip influxdb-protocol because it has been tested separately
in influxdb-protocol.yml
- run: |
- retry() {
- local -i max_attempts=3
- local -i attempt=1
- local -i retry_sleep=5
- local test_output
-
- while [ $attempt -le $max_attempts ]; do
- mvn clean verify \
- -P with-integration-tests \
- -DskipUTs \
- -DintegrationTest.forkCount=1 -DConfigNodeMaxHeapSize=256
-DDataNodeMaxHeapSize=1024 -DDataNodeMaxDirectMemorySize=768 \
- -DClusterConfigurations=${{ matrix.cluster1 }},${{
matrix.cluster2 }},${{ matrix.cluster3 }} \
- -pl integration-test \
- -am -PMultiClusterIT3 \
- -ntp >> ~/run-tests-$attempt.log && return 0
- test_output=$(cat ~/run-tests-$attempt.log)
-
- echo "==================== BEGIN: ~/run-tests-$attempt.log
===================="
- echo "$test_output"
- echo "==================== END: ~/run-tests-$attempt.log
======================"
-
- if ! mv ~/run-tests-$attempt.log
integration-test/target/cluster-logs/ 2>/dev/null; then
- echo "Failed to move log file ~/run-tests-$attempt.log to
integration-test/target/cluster-logs/. Skipping..."
- fi
-
- if echo "$test_output" | grep -q "Could not transfer artifact";
then
- if [ $attempt -lt $max_attempts ]; then
- echo "Test failed with artifact transfer issue, attempt
$attempt. Retrying in $retry_sleep seconds..."
- sleep $retry_sleep
- attempt=$((attempt + 1))
- else
- echo "Test failed after $max_attempts attempts due to
artifact transfer issue."
- echo "Treating this as a success because the issue is likely
transient."
- return 0
- fi
- elif [ $? -ne 0 ]; then
- echo "Test failed with a different error."
- return 1
- else
- echo "Tests passed"
- return 0
- fi
- done
- }
- retry
- - name: Upload Artifact
+ -am \
+ -ntp
+
+ EXIT_CODE=$?
+
+ # 收集日志（每次跑完都把当前的日志存一下，防止 runner 被重置）
+ if [ -d "integration-test/target/cluster-logs" ]; then
+ # 简单的打包一下当前状态，防止文件被 clean 清空
+ tar -czf integration-test/target/last-run-logs-$attempt.tar.gz
integration-test/target/cluster-logs || true
+ fi
+
+ if [ $EXIT_CODE -ne 0 ]; then
+ echo "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
+ echo "FAIL: Test failed on attempt #$attempt"
+ echo "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
+ exit 1
+ else
+ echo "PASS: Attempt #$attempt succeeded. Looping again
immediately..."
+ attempt=$((attempt + 1))
+ # 如果你想让机器歇一会，可以在这加个 sleep 10
+ fi
+ done
+
+ - name: Upload Failure Logs
if: failure()
uses: actions/upload-artifact@v6
with:
- name: cluster-log-triple-java${{ matrix.java }}-${{ runner.os }}-${{
matrix.cluster1 }}-${{ matrix.cluster2 }}-${{ matrix.cluster3 }}
- path: integration-test/target/cluster-logs
- retention-days: 30
+ name: failure-logs-${{ matrix.cluster1 }}-${{ matrix.cluster2 }}
+ path: |
+ integration-test/target/cluster-logs
+ integration-test/target/last-run-logs-*.tar.gz
+ retention-days: 30
\ No newline at end of file
