[Starlingx-discuss] Python2 -> Python3

Sun, Austin austin.sun at intel.com
Thu Jul 4 03:42:48 UTC 2019


Hi Cindy:
    Yes. we will do it and update sheet. 

Thanks.
BR
Austin Sun. 

-----Original Message-----
From: Xie, Cindy [mailto:cindy.xie at intel.com] 
Sent: Thursday, July 4, 2019 11:37 AM
To: Hu, Yong <yong.hu at intel.com>; starlingx-discuss at lists.starlingx.io
Subject: Re: [Starlingx-discuss] Python2 -> Python3

Austin,
Can you add one more column in your xls sheet: https://bugs.launchpad.net/starlingx/+bug/1808073/+attachment/5274592/+files/rpm_python_status-stx.2.0.xlsx:

In your column "I", for your "3rd party" category, to break down "CentOS package" & "3rd party package", so that we can understand how much we can rely on CentOS 8.0, especially for those "risk" ones.

Thanks. - cindy

-----Original Message-----
From: Yong Hu [mailto:yong.hu at intel.com]
Sent: Thursday, July 4, 2019 11:25 AM
To: starlingx-discuss at lists.starlingx.io
Subject: Re: [Starlingx-discuss] Python2 -> Python3

A general question, do we have experience and confidence to ensure 2 python versions (both interpreters and pip libs) can co-exist and each of packages can correctly refer to them??

In my view the best solution is to wait for CentOS 8.0 :-)


On 03/07/2019 2:55 PM, Dean Troyer wrote:
> On 7/3/19 4:07 PM, Saul Wold wrote:
>> The current proposal seems to be to completely convert the base
>> CentOS7.6 system level python to use python3, this carries a high 
>> risk factor as changing out all system-level python code could have a 
>> cascade effect on system functionality and additional dependencies.
>> While
> 
> Changing the distro/system Python version out from under the rest of 
> the distro seems like an enormous time sink, much less a significant 
> reliability risk.
> 
>> A better solution would be to build python3 and the associated 
>> requirements from the existing RHEL EPEL (Extra Packages for 
>> Enterprise Linux) Source RPMs repo and install them into the ISO.
>> This version correctly installs in a segregated directory tree.
> 
> We would probably want to run a significant subset of the upstream 
> OpenStack testing on this combination as it is not (AFAIK) tested there.
>   But this is true of any runtime + distro combination that is not in 
> the fairly short list of combinations that upstream OpenStack actively 
> tests.
> 
>> Another option would be to delay the actual python2 conversion to 
>> StarlingX 4.0, the OpenStack Train release will still support python2.
> 
> One downside to this is it leaves us no margin to defer the change 
> again, this is our second chance as it were.  OpenStack U (as of now) 
> is likely to drop py2 support as a guarantee across-the-board.
> 
>> There is still work that is needed beyond the conversion of the 
>> python code itself to things like RPM specfiles data and other source 
>> code (such as, C code that has #includes of python2.7). It's not 
>> clear to me how much functional testing with python3 has occurred for 
>> the flock beyond what Dean has started with devstack.
> 
> I managed to get the fault services running on py3, sysinv fell over 
> during the dbsync in my quick post-PTG trial run.  That is as far as I 
> took it.  Anyone who wants to try can pick out the local.conf I posted 
> [0]
> 
> dt
> 
> [0] http://paste.openstack.org/show/753844/
> 

_______________________________________________
Starlingx-discuss mailing list
Starlingx-discuss at lists.starlingx.io
http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________
Starlingx-discuss mailing list
Starlingx-discuss at lists.starlingx.io
http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss


More information about the Starlingx-discuss mailing list