Hello Jacek,
Of course, I agree that the code can't be frozen for 2.5 months, but i
think my initial proposal fit with this constraint.
Furthermore, i think that the CC-in2p3 sysadmins needs to have a quite
stable installation interface for Qserv as sooner as possible.
In order what they will learn in the first installations step will be
mostly re-used in the next steps.
Furthermore the LPC team is able to maintain the straightforward
installation procedure during the whole installation process at
CC-in2p3, and i think this procedure can be very easilly scripted in
Puppet (it may require only a few lines).
So, i propose you to integrate this procedure and, if possible, the
tests on small dataset in the next release candidate, that we could
deliver in two weeks for example.
In the next releases coming before July, we will make everything
possible to keep the installation procedure the same, and the LPC team
could improve it w.r.t. CC-in2p3 sysadmin advices, and make it compliant
with development features. This plan would allow the CC-in2p3 sysadmins
to prepare the 300 nodes installation with confidence and with an good
interaction with development team.
We can talk about that on Thursday. What time do you prefer to call ?
Around 9h PM pacific time would be perfect for us.
Thanks again for your help,
Fabrice
On 04/08/2013 07:02 PM, Jacek Becla wrote:
> Fabrice
>
> We have to be more agile, we are still working on features
> that are essential for the large scale test, we can't
> freeze the code 2 1/2 months before the test. I expect
> we will be merging the new changes and fixes into the master
> through end of May or so, and leave 2-3 weeks for final
> validation.
>
> I hope you will be able to accommodate this. Of course
> we will try to avoid disrupting the puppet scripts as
> much as we can.
>
> p.s. I now have the metadata integrated with (some parts of)
> qsers, (#2725), I expect to have that reviewed and in the
> master shortly.
>
> Perhaps let's discuss details this Thursday? I'll try to
> call in.
>
>
> Thanks,
> Jacek
>
>
>
> On 04/08/2013 06:09 AM, Fabrice Jammes wrote:
>> Hello,
>>
>> In order to ease the installation of 300 nodes cluster at in2p3, we
>> propose this work plan for next two weeks :
>>
>> 1. solve the integration of new mysqld-based export path support (i.e.
>> update install script in order to support this feature, in mono and
>> multi-nodes) : #2796, and merge to master : 5 to 8 days for LPC team
>> 2. merge the 2014 ticket containing the utomated tests on small data
>> sets (it has already been reviewed, but it doesn't work after the merge,
>> due to 1.) : 1 to 4 days for LPC team
>> 3. make a release with : the automated install procedure (already merge
>> after out stay at SLAC), the tests on small dataset, and the
>> integration with new mysqld-based export path support : would you agree
>> to do this ?
>> The CC-IN2P3 sysadmin needs to have this release in order to start
>> the integration inside Puppet. The install procedure is documented in
>> README.txt of the master branch.
>>
>> We think that all these features will make simpler the install and the
>> admin of the 300 nodes cluster, it could also be very useful for
>> continuous integration with buildbot (we think it will useful not only
>> to compile Qserv but also run the test suite).
>>
>> Would you agree with this ?
>>
>> The LPC team.
>>
>> ########################################################################
>> Use REPLY-ALL to reply to list
>>
>> To unsubscribe from the QSERV-L list, click the following link:
>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
>>
>
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the QSERV-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
|