kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-14 06:27:45 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

[None][ci] Increase the number of retries in docker image generation (#7557)

Browse Source 
Signed-off-by: Yanchao Lu <yanchaol@nvidia.com>
This commit is contained in:
Yanchao Lu 2025-09-05 14:47:14 +08:00 committed by GitHub
parent 8e3962d278
commit 4195010e13
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
jenkins/BuildDockerImage.groovy
View File

@ -295,8 +295,8 @@ def buildImage(config, imageKeyToTag)
if (dependent) {
stage ("make ${dependent.target}_${action} (${arch})") {
def randomSleep = (Math.random() * 300 + 300).toInteger()
trtllm_utils.llmExecStepWithRetry(this, script: "docker pull ${TRITON_IMAGE}:${TRITON_BASE_TAG}", sleepInSecs: randomSleep, shortCommondRunTimeMax: 7200)
def randomSleep = (Math.random() * 600 + 600).toInteger()
trtllm_utils.llmExecStepWithRetry(this, script: "docker pull ${TRITON_IMAGE}:${TRITON_BASE_TAG}", sleepInSecs: randomSleep, numRetries: 6, shortCommondRunTimeMax: 7200)
trtllm_utils.llmExecStepWithRetry(this, script: """
cd ${LLM_ROOT} && make -C docker ${dependent.target}_${action} \
BASE_IMAGE=${BASE_IMAGE} \
@ -305,7 +305,7 @@ def buildImage(config, imageKeyToTag)
IMAGE_WITH_TAG=${dependentImageWithTag} \
STAGE=${dependent.dockerfileStage} \
BUILD_WHEEL_OPTS='-j ${build_jobs}' ${args}
""", sleepInSecs: randomSleep, numRetries: 3, shortCommondRunTimeMax: 7200)
""", sleepInSecs: randomSleep, numRetries: 6, shortCommondRunTimeMax: 7200)
args += " DEVEL_IMAGE=${dependentImageWithTag}"
if (target == "ngc-release") {
imageKeyToTag["NGC Devel Image ${config.arch}"] = dependentImageWithTag
@ -324,8 +324,8 @@ def buildImage(config, imageKeyToTag)
}
stage ("make ${target}_${action} (${arch})") {
sh "env | sort"
def randomSleep = (Math.random() * 300 + 300).toInteger()
trtllm_utils.llmExecStepWithRetry(this, script: "docker pull ${TRITON_IMAGE}:${TRITON_BASE_TAG}", sleepInSecs: randomSleep, shortCommondRunTimeMax: 7200)
def randomSleep = (Math.random() * 600 + 600).toInteger()
trtllm_utils.llmExecStepWithRetry(this, script: "docker pull ${TRITON_IMAGE}:${TRITON_BASE_TAG}", sleepInSecs: randomSleep, numRetries: 6, shortCommondRunTimeMax: 7200)
trtllm_utils.llmExecStepWithRetry(this, script: """
cd ${LLM_ROOT} && make -C docker ${target}_${action} \
BASE_IMAGE=${BASE_IMAGE} \
@ -334,7 +334,7 @@ def buildImage(config, imageKeyToTag)
IMAGE_WITH_TAG=${imageWithTag} \
STAGE=${dockerfileStage} \
BUILD_WHEEL_OPTS='-j ${build_jobs}' ${args}
""", sleepInSecs: randomSleep, numRetries: 3, shortCommondRunTimeMax: 7200)
""", sleepInSecs: randomSleep, numRetries: 6, shortCommondRunTimeMax: 7200)
if (target == "ngc-release") {
imageKeyToTag["NGC Release Image ${config.arch}"] = imageWithTag
}
@ -686,7 +686,7 @@ pipeline {
}
cmd += imageKeyToTag.values().join(" ")
withCredentials([usernamePassword(credentialsId: "NSPECT_CLIENT-${nspect_env}", usernameVariable: 'NSPECT_CLIENT_ID', passwordVariable: 'NSPECT_CLIENT_SECRET')]) {
trtllm_utils.llmExecStepWithRetry(this, script: cmd, numRetries: 3, shortCommondRunTimeMax: 7200)
trtllm_utils.llmExecStepWithRetry(this, script: cmd, numRetries: 6, shortCommondRunTimeMax: 7200)
}
}
}