Thanks Evans. 

Cos: We do have the option to make new point releases and convenience repos 
from older code if someone asks, but I suspect there won't be such demand. 


> On Oct 20, 2015, at 5:25 PM, Konstantin Boudnik <[email protected]> wrote:
> 
> Another part here is that _all_ our historical releases are gone ;(
> 
> It might be not a huge issue as we clearly encourage our users to stay on the
> later stuff, but still it's a blow to the project. One of those non-so-obvious
> things that are very clear when looked upon in the rear-view mirror.
> 
> Cos
> 
>> On Wed, Oct 21, 2015 at 02:34AM, Evans Ye wrote:
>> Thanks Andrew!
>> For temporarily dev/test usage, I've built packages back on our CI server:
>> 
>> http://ci.bigtop.apache.org:8080/view/Releases/
>> 
>> And added some wiki for users:
>> https://cwiki.apache.org/confluence/display/BIGTOP/How+to+install+Bigtop+1.0.0+with+Bigtop+Provisioner
>> 
>> I think this is enough, if it won't take too long to get bigtop bucket back
>> online.
>> 
>> OTOH, BIGTOP-2092 <https://issues.apache.org/jira/browse/BIGTOP-2092> reveals
>> that all the historical release artifacts are not available as well.
>> My thought is to just provide 1.0 release artifacts.
>> 
>> Cos,
>> right now I haven't used up the resource provided by Tom for CI. If needed
>> I think we can re-negotiate with Tom for rearrangement to get some S3
>> resources.
>> Anyhow, if you need help, just ping me.
>> 
>> 
>> 
>> 
>> 
>> 
>> 2015-10-21 0:22 GMT+08:00 Andrew Purtell <[email protected]>:
>> 
>>> FWIW, I was able to create the following buckets under my account:
>>> 
>>> bigtop-repo
>>> bigtop-repos
>>> 
>>> If you want to go with one of these, even if only temporarily, let me know
>>> and I'll send the PMC access credentials of an IAM user with full perms
>>> over these buckets.
>>> 
>>> 
>>> 
>>> On Mon, Oct 19, 2015 at 2:08 PM, Sean Mackrory <[email protected]>
>>> wrote:
>>> 
>>>>>> Sean, could you figure out how this can be done?
>>>> 
>>>> Sure. Although if the bucket ceased to exist when the account was
>>>> terminated, the name should have been free again with a couple of hours.
>>> So
>>>> either the bucket has NOT ceased to exist, or someone other than Cloudera
>>>> now has the bucket name.
>>>> 
>>>> I'll find out which...
>>>> 
>>>> On Mon, Oct 19, 2015 at 12:56 PM, Konstantin Boudnik <[email protected]>
>>>> wrote:
>>>> 
>>>>>> On Sun, Oct 18, 2015 at 03:30PM, Evans Ye wrote:
>>>>>> Guys I've a bad news.
>>>>>> I'm guessing that our official released 1.0 repos on S3 are also tied
>>>>> with
>>>>>> Cloudera's credentials, which is why we no longer have S3 repos
>>>> available
>>>>>> for users to consume now(I've tried centos6 and debian8).
>>>>>> That's really bad in user experience. :(
>>>>>> Cos,
>>>>>> * Could you please confirm where we put 1.0 repos on?
>>>>>> * If my guess is right, do you still have copies of signed repos at
>>>> local
>>>>>> for restore?
>>>>> 
>>>>> I do have the copy of the repos, so it should be an easy exercise to
>>>>> restore
>>>>> them to the new location.
>>>>> 
>>>>>> * I don't have knowledge of how our current S3 resources are being
>>>>> managed.
>>>>>> And we don't have S3 resources available from Tom's team, hence I
>>> think
>>>>> we
>>>>>> need to plan for this now.
>>>>>> 
>>>>>> Evans
>>>>> 
>>>>>> On Sun, Oct 18, 2015 at 10:53AM, Andrew Purtell wrote:
>>>>>> Since the binary convenience artifacts are not an official release
>>>>> artifact,
>>>>>> only the source tarball is, then any of us can feel free to use the
>>>>> official
>>>>>> 1.0 release tarball to generate a new set of packages, store them at
>>>> new
>>>>>> locations, and update pointers to that location.
>>>>>> 
>>>>>> The new location *could* be Apache dist. Other projects host their
>>>>>> convenience artifacts there. We need to consider the impact on
>>>>>> infrastructure costs. I don't think there would be any significant
>>>>> impact.
>>>>>> We could mail infrastructure to find out if they have any concerns
>>>> given
>>>>> the
>>>>>> space requirement if you prefer this option.
>>>>> 
>>>>> We have discussed it with INFRA in the early days of the project. And
>>>>> precisely the cost impact was the reason we have kept it elsewhere. We
>>>> are
>>>>> talking about ~0.7GB/repo x 5 platforms (at least) x number of Apache
>>>>> mirrors
>>>>> - it's pretty huge, really.
>>>>> 
>>>>>> Bintray is another option. I don't know anything about it.
>>>>>> 
>>>>>> I also looked at creating a S3 bucket for Bigtop using my account. I
>>>>> have a
>>>>>> feeling monthly transfer charges will not be a problem. However the
>>>>> 'bigtop'
>>>>>> bucket is taken. Perhaps we could talk to Tom about getting ownership
>>>>>> transferred if you prefer this option.
>>>>> 
>>>>> We need to ask Cloudera's infra team to transfer it to us. Sean, could
>>>> you
>>>>> figure out how this can be done?
>>>>> 
>>>>> Thanks!
>>>>>  Cos
>>>>> 
>>>>>>> On Oct 18, 2015, at 4:42 AM, Olaf Flebbe <[email protected]> wrote:
>>>>>>> 
>>>>>>> Hi Cos,
>>>>>>> 
>>>>>>> Is it correct that we should not put our release artifacts on
>>> apache
>>>>> dist,
>>>>>>> since Apache is about code not binaries?
>>>>>>> 
>>>>>>> Can we use something like bintray.org ? Looks like the
>>> functionality
>>>>> we need: RPM and DEB repo, and does have an API.
>>>>>>> 
>>>>>>> Olaf
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>>> Am 18.10.2015 um 09:30 schrieb Evans Ye <[email protected]>:
>>>>>>>> 
>>>>>>>> Guys I've a bad news.
>>>>>>>> I'm guessing that our official released 1.0 repos on S3 are also
>>>> tied
>>>>> with
>>>>>>>> Cloudera's credentials, which is why we no longer have S3 repos
>>>>> available
>>>>>>>> for users to consume now(I've tried centos6 and debian8).
>>>>>>>> That's really bad in user experience. :(
>>>>>>>> Cos,
>>>>>>>> * Could you please confirm where we put 1.0 repos on?
>>>>>>>> * If my guess is right, do you still have copies of signed repos
>>> at
>>>>> local
>>>>>>>> for restore?
>>>>>>>> * I don't have knowledge of how our current S3 resources are being
>>>>> managed.
>>>>>>>> And we don't have S3 resources available from Tom's team, hence I
>>>>> think we
>>>>>>>> need to plan for this now.
>>>>>>>> 
>>>>>>>> Evans
>>> 
>>> 
>>> 
>>> --
>>> Best regards,
>>> 
>>>   - Andy
>>> 
>>> Problems worthy of attack prove their worth by hitting back. - Piet Hein
>>> (via Tom White)
>>> 

Reply via email to