Martijn Visser created FLINK-40058:
--------------------------------------
Summary: cron_snapshot_deployment nightly jobs run out of disk
space on hosted agents
Key: FLINK-40058
URL: https://issues.apache.org/jira/browse/FLINK-40058
Project: Flink
Issue Type: Bug
Components: Build System / Azure Pipelines, Build System / CI
Reporter: Martijn Visser
Assignee: Martijn Visser
The nightly cron_snapshot_deployment_binary and cron_snapshot_deployment_maven
jobs emit repeated "Free disk space on / is lower than 5%" warnings. In build
76675 usage peaked at 99.18% (binary) and 97.35% (maven), so the jobs are one
dependency bump away from failing with "no space left on device".
Both jobs run on Microsoft-hosted ubuntu-24.04 agents, which start with only
~16 GB free. On top of that the jobs pull the multi-JDK flink-ci-docker image,
restore a 4.2 GB Maven cache, run a near-full Flink build (-pl flink-dist -am),
and produce the extracted dist plus tgz copies.
Every other hosted-agent job (jobs-template.yml, e2e-template.yml) already runs
tools/azure-pipelines/free_disk_space.sh first, freeing ~34 GB;
build-nightly-dist.yml is the one hosted-agent template that never got this
step.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
