[ 
https://ovirt-jira.atlassian.net/browse/OVIRT-1005?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Barak Korren reassigned OVIRT-1005:
-----------------------------------

    Assignee: Barak Korren  (was: infra)

> Fix race condition caused by experimental cleanup script
> --------------------------------------------------------
>
>                 Key: OVIRT-1005
>                 URL: https://ovirt-jira.atlassian.net/browse/OVIRT-1005
>             Project: oVirt - virtualization made easy
>          Issue Type: Improvement
>            Reporter: Barak Korren
>            Assignee: Barak Korren
>            Priority: Highest
>
> The experimental repo has a cleanup script that is ran every hour via cron. 
> The script is meant to delete old and unused package builds.
> The content of the script is as follows:
> {code}
> find \
>     /srv/resources/repos/ovirt/experimental/*/* \
>     -maxdepth 0 \
>     -type d \
>     -not -name 'latest.*' \
>     -and \
>     -mtime +1 \
>     -exec rm -rf {} \;
> {code}
> The problem here is that 'latest' is not filtered out to not be deleted by 
> this script. It is OK to delete it in most cases because then it gets 
> recreated by the 'deploy-to-experimental' job, but there might be a case 
> where 'test-repo_ovirt-experimental_*' can run before this happens, this may 
> cause it to get an empty or partial 'latest' repo.
> To remedy this, the script needs to be changed to be the following:
> {code}
> find \
>     /srv/resources/repos/ovirt/experimental/*/* \
>     -maxdepth 0 \
>     -type d \
>     -not -name 'latest.*' \
>     -and \
>     -not -name 'latest' \
>     -and \
>     -mtime +1 \
>     -exec rm -rf {} \;
> {code}
> The script itself on the resources server was already changed in-palce, this 
> ticket is so we  this change is also "backed up" to the 'infra-puppet' repo.



--
This message was sent by Atlassian JIRA
(v1000.656.2#100024)
_______________________________________________
Infra mailing list
[email protected]
http://lists.ovirt.org/mailman/listinfo/infra

Reply via email to