diff --git a/jenkins/L0_Test.groovy b/jenkins/L0_Test.groovy index d71d510e36c..77335ac3415 100644 --- a/jenkins/L0_Test.groovy +++ b/jenkins/L0_Test.groovy @@ -1,4 +1,4 @@ -@Library(['bloom-jenkins-shared-lib@main', 'trtllm-jenkins-shared-lib@main']) _ +@Library(['bloom-jenkins-shared-lib@dev-yuanjingx-add_custom_ssh_option', 'trtllm-jenkins-shared-lib@main']) _ import java.lang.InterruptedException import groovy.transform.Field @@ -846,7 +846,7 @@ def runLLMTestlistWithSbatch(pipeline, platform, testList, config=VANILLA_CONFIG sh "cd ${llmPath} && tar -zxf ${BUILD_CONFIGS[config][TARNAME]}" // Upload slurm_run_sh to Frontend node - def scriptRunLocalPath = "${llmSrcLocal}/jenkins/scripts/slurm_run.sh" + def scriptRunLocalPath = "${llmSrcLocal}/jenkins/scripts/slurm_run_test.sh" Utils.exec(pipeline, script: "echo \"Script to trigger slurm job: \" && cat ${scriptRunLocalPath}") Utils.copyFileToRemoteHost( @@ -997,10 +997,11 @@ def runLLMTestlistWithSbatch(pipeline, platform, testList, config=VANILLA_CONFIG stage("[${stageName}] Run Pytest") { Utils.exec( pipeline, - timeout: false, + timeout: 200, script: Utils.sshUserCmd( remote, - scriptExecPathNode + scriptExecPathNode, + "-o ConnectTimeout=600" ) ) } diff --git a/jenkins/scripts/slurm_run_test.sh b/jenkins/scripts/slurm_run_test.sh new file mode 100755 index 00000000000..cae084337e9 --- /dev/null +++ b/jenkins/scripts/slurm_run_test.sh @@ -0,0 +1,4 @@ +#!/bin/bash +nvidia-smi +sleep 1000 +echo "Done"