[
https://issues.apache.org/jira/browse/MNG-6465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16591585#comment-16591585
]
Lasse Westh-Nielsen commented on MNG-6465:
------------------------------------------
[~rfscholte] we already have a proxy cache - same deal. As soon as requests
start crossing the network, there is a small failure rate. Multiplied to our
scale, it is an unacceptable failure rate.
So effectively there is Maven client going to our proxy cache, which in turns
goes to Maven Central CDN - and _still_ failure rates are too high. Our proxy
cache improved things and we got more throughput (less builds failing due to
artifact download errors) but it is still not good enough.
So in my view this is a client issue: I need my Maven client to behave more
leniently, taking more retries and perhaps longer timeouts to ensure better
throughput. And I am happy to pay with reduced latency.
So is this stuff configurable or not?
> Artifact download failures
> --------------------------
>
> Key: MNG-6465
> URL: https://issues.apache.org/jira/browse/MNG-6465
> Project: Maven
> Issue Type: Bug
> Components: Artifacts and Repositories
> Affects Versions: 3.5.0
> Environment: Windows, Ubuntu
> Reporter: Lasse Westh-Nielsen
> Priority: Major
>
> Hey,
> We often run Maven builds of [https://github.com/neo4j/neo4j] on machines
> with cold caches/ empty ~/.m2/repository folders.
> There are rather a lot of dependencies to download, so even though the
> failure rate of individual artifact downloads is small, at our scale it
> causes significant build failures.
> I am working on seeding caches. Separately though it would be great to be
> able to configure retries and timeouts for dependency downloads - optimise
> for build throughput rather than latency as it were.
> How do I do that? I can't find it documented anywhere, and trying to browse
> "wagon" source code got me nowhere.
> Regards,
> Lasse
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
