[OpenStack-Infra] Please delete tag 2.0.1
Scott Little
scott.little at windriver.com
Fri Oct 18 18:10:25 UTC 2019
We'll create a 2.0.1b tag for the manifest.
We withdraw or request for tag deletion.
Thanks, Scott
On 2019-10-17 5:15 p.m., Jeremy Stanley wrote:
> [Cc'ing because it seems you're not subscribed, apologies for any
> duplicate copies you may receive]
>
> On 2019-10-17 15:14:38 -0400 (-0400), Scott Little wrote:
>> Tag 2.0.1 should point to c89aeff3401fa37ca679bf170cd2f239a51b639f in repo
>> opendev.org/starlingx/manifest .
>>
>> The current 2.0.1 tag was pushed in error. Please delete the 2.0.1 tag from
>> opendev.org/starlingx/manifest so that I can push the correct tag.
>>
>> We realize the challenges of re-writing a tag, and will be informing our
>> users of the change and the manual steps to fix their repos.
> To recap the discussion from #openstack-infra on Freenode, doing
> this will leave Zuul and Nodepool repository caches in an
> inconsistent state from the official copy in Gerrit due to Git not
> communicating changes for tags. While it sounds like the CI jobs
> currently applied to or relying on the starlingx/manifest repository
> do not take any action on nor generate any artifacts from its Git
> tags, we would still need to perform cleanup under maintenance to
> wipe and replace Git caches on the following servers for
> consistency:
>
> 12 Zuul executors
> 8 Zuul mergers
> 3 Nodepool image builders
>
> This must be done with their respective services offline to force
> them to re-clone the repositories without erroneous tags when they
> start. Just as a reminder, these systems are designed with the
> assumption that Git tags won't be deleted or replaced because Git
> itself provides no mechanism for a client to discover that this has
> happened.
>
> Given the amount of manual effort involved, please consider
> alternatives such pushing a tag with a new (slightly higher) version
> number. While deleting that tag is something we *can* do, we need to
> know that the effort is justified, and not simply the OpenDev
> sysadmins taking on more work to save someone else from doing
> (perhaps a smaller amount of) work instead.
More information about the OpenStack-Infra
mailing list