[
https://issues.apache.org/jira/browse/CASSANDRA-21403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18086425#comment-18086425
]
Michael Semb Wever commented on CASSANDRA-21403:
------------------------------------------------
Small additions to the Jenkinsfile
- to fix retries (which was no longer working when agents/nodes died, since
CASSANDRA-20833) (this is visible in the ci builds)
- @NonCPS where (now) needed
- fix default value for the branch parameter (after a new helm deploy)
Small additions to .build/run-ci
- fix missing parameters on jobs also at helm deploy time
- .git suffixes on repo urls are optional
Small additions to .jenkins/k8s/
- update docs for agents small resource requirements (needed a small bump)
- add the cassandra-6.0 job
- enforce a limit of one agent pod per node for simpler operations (like
orphan cleanup)
CI
- 5.0 ci-cassandra.infra.datastax.com/job/cassandra-5.0/148/
- 6.0 ci-cassandra.infra.datastax.com/job/cassandra-6.0/25/
- trunk http://ci-cassandra.infra.datastax.com/job/cassandra/146/
> Preload maven dependencies and gradle binary in docker build/test images
> ------------------------------------------------------------------------
>
> Key: CASSANDRA-21403
> URL: https://issues.apache.org/jira/browse/CASSANDRA-21403
> Project: Apache Cassandra
> Issue Type: Improvement
> Components: Build, CI
> Reporter: Michael Semb Wever
> Assignee: Michael Semb Wever
> Priority: Normal
> Fix For: 5.0.x, 6.x, 7.x
>
> Attachments: ci_summary_thelastpickle_mck-21403-5.0_112.html,
> ci_summary_thelastpickle_mck-21403-6.0_13.html,
> ci_summary_thelastpickle_mck-21403-trunk_93.html,
> results_details_thelastpickle_mck-21403-5.0_112.tar.xz,
> results_details_thelastpickle_mck-21403-6.0_13.tar.xz,
> results_details_thelastpickle_mck-21403-trunk_93.tar.xz
>
> Time Spent: 2h
> Remaining Estimate: 0h
>
> Maven dependencies (and the gradle binary for accord's gradle wrapper) were
> being downloaded on every CI pipeline run. This lead to hitting rate limits
> and runs failing.
> The docker containers mount the host's ~/.m2/repository to try to reduce this
> download cost, but CI can't do this for two reasons:
> 1. pre-ci.cassandra.a.o spins up new agents each build
> 2. ci-cassandra.a.o mounts a tmp directory, as it runs multiple executors on
> the same agent hosts, and multiple containers cannot use the same repository
> directory (or files become corrupt).
> The solution is to bake in the maven dependencies and gradle binary into the
> images, and rsync them into place when the container runs.
> The preloaded set of dependencies will drift over time as dependencies are
> added or versions bumped, but this is expected to be a minor issue remedied
> in time when the image is built and deployed again. (It can be freely rebuilt
> at any time too.)
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
