Hi Evans,

a) On a second thought: We are working on the overlay filesystem which is known 
not to be following 100% POSIX compliance:
https://docs.docker.com/storage/storagedriver/overlayfs-driver/#limitations-on-overlayfs-compatibility
 
<https://docs.docker.com/storage/storagedriver/overlayfs-driver/#limitations-on-overlayfs-compatibility>
Maybe the build/lock problems are related.

b) IIRC is more or less by design and can't be fixed. 

Maybe it was a-not-so-good approach suggested. So I would vote to revoke 
BIGTOP-3001, rebuild images and revert back to old build process.


Olaf


> Am 01.07.2018 um 18:57 schrieb Evans Ye <[email protected]>:
> 
> As I approaching the new way to build packages via bigtop-ci, I discover
> some issues a) gradle home locked, hence can't build in parallel. b) many
> dangling images are generated.
> 
> Though those can be tacked, I think we better revert BIGTOP-3001 first and
> fix those issues one by one.
> Any second thought?
> 
> 
> Evans Ye <[email protected]> 於 2018年6月14日 週四 下午10:28寫道:
> 
>> Well then we can give bigtop-ci a try.
>> I've integrated it into a testing job:
>> 
>> 
>> https://ci.bigtop.apache.org/view/Packages/job/Bigtop-trunk-packages-BIGTOP-2993/
>> 
>> Let me see how it goes.
>> 
>> 
>> Olaf Flebbe <[email protected]> 於 2018年6月14日 週四 下午11:23寫道:
>> 
>>> hi evans
>>> 
>>> i disabled nexus amteady before since it had issues as well. so no nexus
>>> should not be a problem
>>> 
>>> olag
>>> 
>>>> Am 14.06.2018 um 15:54 schrieb Evans Ye <[email protected]>:
>>>> 
>>>> bigtop-ci/build.sh can solve the issue but right now we don't have nexus
>>>> enabled with it. So it's not ready as Olaf said. I think reverting is a
>>>> better option since we don't want to loose stability after day-to-day
>>>> iterations.
>>>> We can recommit BIGTOP-3001 back once bigtop-ci is polished and ready to
>>>> take over CI jobs.
>>>> 
>>>> Evans
>>>> 
>>>> 
>>>> Olaf Flebbe <[email protected]> 於 2018年6月14日 週四 下午10:05寫道:
>>>> 
>>>>> Hi,
>>>>> 
>>>>> there was a Change in bigtop to remap the uid of jenkins to uid 500.
>>> Now
>>>>> our CI is seriously broken, since this uid is actually reservered by
>>> our
>>>>> cloud provider for our host and the jenkins is now broken: since
>>> jenkins on
>>>>> the host and within the container has different uid's  . Mapping of
>>> sources
>>>>> into the container do not work any more.
>>>>> 
>>>>> I already had to remove all the data in Bigtop-packages-trunk CI job.
>>>>> 
>>>>> There are a couple of options to solve this:
>>>>> 
>>>>> * Reverting BIGTOP-3001, but than it may clash with uid at the
>>> committers
>>>>> machine
>>>>> * Enhance bigtop-ci/build.sh
>>>>> 
>>>>> Best,
>>>>> Olaf
>>> 
>> 

Reply via email to