From 19dff5f3703e8b4721cd722ea763ddd36b467b0d Mon Sep 17 00:00:00 2001 From: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> Date: Wed, 13 Aug 2025 00:49:35 -0700 Subject: [PATCH 1/3] [TRTLLM-7169][infra] Fix slurm test failed will show "Submit Test Results"/"Checking if the Node is Online" in test name Signed-off-by: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> --- jenkins/L0_Test.groovy | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/jenkins/L0_Test.groovy b/jenkins/L0_Test.groovy index 4565193c349e..eb8f4cf50658 100644 --- a/jenkins/L0_Test.groovy +++ b/jenkins/L0_Test.groovy @@ -131,11 +131,13 @@ def uploadResults(def pipeline, SlurmCluster cluster, String nodeName, String st "results-${stageName}.tar.gz", "${UPLOAD_PATH}/test-results/" ) - junit(testResults: "${stageName}/results*.xml") } else { println("No results xml to submit") } } + if (downloadSucceed) { + junit(testResults: "${stageName}/results*.xml") + } } } From 40a85286d7a06e6687fce2cba69fd4dc3f2407d9 Mon Sep 17 00:00:00 2001 From: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> Date: Wed, 13 Aug 2025 00:55:33 -0700 Subject: [PATCH 2/3] Fix Signed-off-by: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> --- jenkins/L0_Test.groovy | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/jenkins/L0_Test.groovy b/jenkins/L0_Test.groovy index eb8f4cf50658..7bc6d358f94d 100644 --- a/jenkins/L0_Test.groovy +++ b/jenkins/L0_Test.groovy @@ -425,6 +425,10 @@ def runLLMTestlistOnSlurm(pipeline, platform, testList, config=VANILLA_CONFIG, p } else { error "The Slurm node does not come online in the waiting period. Terminating the job." } + slurmRunner = runInDockerOnNodeMultiStage(LLM_DOCKER_IMAGE, nodeName, dockerArgs, false) + executeLLMTestOnSlurm(pipeline, platform, testList, config, perfMode, stageName, splitId, splits, skipInstallWheel, cpver, slurmRunner) + } else { + echo "The node does not come online in 2 hours, terminating the job" } } finally { stage("Clean up SLURM Resources") { From b8a9fdcc697660c4f7f5cb9701b49d45bef9b284 Mon Sep 17 00:00:00 2001 From: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> Date: Mon, 1 Sep 2025 02:56:33 -0700 Subject: [PATCH 3/3] Fix slurm test format Signed-off-by: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com> --- jenkins/L0_Test.groovy | 4 ---- 1 file changed, 4 deletions(-) diff --git a/jenkins/L0_Test.groovy b/jenkins/L0_Test.groovy index 7bc6d358f94d..eb8f4cf50658 100644 --- a/jenkins/L0_Test.groovy +++ b/jenkins/L0_Test.groovy @@ -425,10 +425,6 @@ def runLLMTestlistOnSlurm(pipeline, platform, testList, config=VANILLA_CONFIG, p } else { error "The Slurm node does not come online in the waiting period. Terminating the job." } - slurmRunner = runInDockerOnNodeMultiStage(LLM_DOCKER_IMAGE, nodeName, dockerArgs, false) - executeLLMTestOnSlurm(pipeline, platform, testList, config, perfMode, stageName, splitId, splits, skipInstallWheel, cpver, slurmRunner) - } else { - echo "The node does not come online in 2 hours, terminating the job" } } finally { stage("Clean up SLURM Resources") {