[jira] [Comment Edited] (MNG-6474) Maven intermittently hanging at end of multithreaded build

["Michael Osipov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/MNG-6474?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16861156#comment-16861156 ] 

Michael Osipov edited comment on MNG-6474 at 6/11/19 7:07 PM:
--------------------------------------------------------------

If you have two jobs (VMs) accessing the same local repo than you are in trouble. The repo impl is not concurrent-safe. Use Takari's implementation for that or distinct repos.


was (Author: michael-o):
If you have two jobs (VM) accessing the same local repo than you are in trouble. The repo impl is not concurrent-safe. Use Takari's implementation for that or distinct repos.

> Maven intermittently hanging at end of multithreaded build
> ----------------------------------------------------------
>
>                 Key: MNG-6474
>                 URL: https://issues.apache.org/jira/browse/MNG-6474
>             Project: Maven
>          Issue Type: Bug
>    Affects Versions: 3.5.2
>            Reporter: Beirtí Ó'Nunáin
>            Priority: Major
>         Attachments: Trunk_Build_after_every_commit_UNDEFINED-r15120.log, thread dump.txt
>
>
> I can't reproduce this with a small skeleton project. It has only started happening since we upgraded from 3.3 to 3.5.
>  
> I have a multithreaded multi-module build which builds a number of large (300mb+) artifacts upon SVN commits. I'm running the following goals:
>  
> mvn clean deploy -T 1C
>  
> Reasonably often, perhaps 1 in 10 builds, the build hangs just before where I'd expect the build to have finished. I'm guessing that maven is attempting to install an artifact to the local repository (looking at the stack trace) while the build is about to complete as when I compare the build log from the hung build to a successful build, the following line is the "Reactor Summary"
> This issue happens both on my local machine  and on our TeamCity server (Both multi-processor, multi-core 14 vs 32 cores)
> The Thread dump isn't indicating any kind of DeadLock. Our sysadmins have checked the box for any antivirus scanning which may be impacting the file copy indicated in the log but can't see anything going on at that time.
> Full thread dump from TeamCity attached
>  



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)