Re: [Stretch] Status for architecture qualification

2016-06-27 Thread Luca Filipozzi
On Mon, Jun 27, 2016 at 04:35:03PM +0200, Wouter Verhelst wrote:
> (sorry for jumping in late here)
> 
> On Wed, Jun 15, 2016 at 07:51:55AM +0800, Paul Wise wrote:
> > On Wed, 2016-06-15 at 01:37 +0300, Dimitri John Ledkov wrote:
> > 
> > > At the openmainframeproject EU meetup, it was indicated that SUSE
> > > joined with indication that Open Build Service might be able to use
> > > resources hosted by Marist.
> > > 
> > > I wonder if it makes sense to reach out, and see if there are
> > > resources available to use as porter boxes & build boxes. That way
> > > Debian might be able to get such donated resource available on ongoing
> > > basis and hopefully with some hw support.
> > 
> > Marist already support Debian with an s390x buildd:
> > 
> > ldapsearch -LLL -x -h db.debian.org -b ou=hosts,dc=debian,dc=org  
> > "(sponsor=*marist*)"
> > https://db.debian.org/machines.cgi?host=zani
> > 
> > Our other sponsors for s390 are www.iic.kit.edu and www.zivit.de:
> > 
> > ldapsearch -LLL -x -h db.debian.org -b ou=hosts,dc=debian,dc=org  
> > "(architecture=s390*)" sponsor
> 
> Given that we already seem to have three hardware sponsors for the s390x
> port, is this really a concern?

Our standard for buildd / porterboxen of a released architecture is:
- multiple machines (N + 1, N sufficient to handle the buildd / porter load)
- under warranty or post-warranty hardware support for the duration of their
  use as buildds / porterboxen including through the LTS timeframe
- located in multiple geographic locations (EU and NA, ideally)
- hosted by different hosting partners, each providing high availability
  (power, cooling, networking) and intelligent remote hands
- under Debian control and/or ownership; available & affordable 
- redundant disks and power supplies
- out-of-band service processor with power management or equivalent

Not satisfying the fifth bullet is a minor concern in the case of s390x.

> If we were to lose one sponsor, we'd still have two (and it would be
> reasonable to try to ensure that we get a new sponsor to replace the one
> lost).

Indeed.  The more redundnant sponsors, the less worrying is the concern.

> How about making it a requirement that there is some level of redundancy
> in sponsors for ports where hardware cannot be easily bought with Debian
> money? That would cover the s390x port as well as any potential other
> insanely-expensive-hardware port[1] that we might support now or in the
> future.

That is our requirement, effectively.  The mild concern has not blocked the
port from releasing.  That said, the concern should be documented.

> If push comes to shove, we could also talk to IBM. Pretty much all POWER
> hardware we have was sponsored by IBM; it's not unreasonable to assume they
> might be willing to also sponsor s390x time or hardware.

-- 
Luca Filipozzi
http://www.crowdrise.com/SupportDebian



Re: [Stretch] Status for architecture qualification

2016-06-27 Thread Wouter Verhelst
(sorry for jumping in late here)

On Wed, Jun 15, 2016 at 07:51:55AM +0800, Paul Wise wrote:
> On Wed, 2016-06-15 at 01:37 +0300, Dimitri John Ledkov wrote:
> 
> > At the openmainframeproject EU meetup, it was indicated that SUSE
> > joined with indication that Open Build Service might be able to use
> > resources hosted by Marist.
> > 
> > I wonder if it makes sense to reach out, and see if there are
> > resources available to use as porter boxes & build boxes. That way
> > Debian might be able to get such donated resource available on ongoing
> > basis and hopefully with some hw support.
> 
> Marist already support Debian with an s390x buildd:
> 
> ldapsearch -LLL -x -h db.debian.org -b ou=hosts,dc=debian,dc=org  
> "(sponsor=*marist*)"
> https://db.debian.org/machines.cgi?host=zani
> 
> Our other sponsors for s390 are www.iic.kit.edu and www.zivit.de:
> 
> ldapsearch -LLL -x -h db.debian.org -b ou=hosts,dc=debian,dc=org  
> "(architecture=s390*)" sponsor

Given that we already seem to have three hardware sponsors for the s390x
port, is this really a concern?

If we were to lose one sponsor, we'd still have two (and it would be
reasonable to try to ensure that we get a new sponsor to replace the one
lost).

How about making it a requirement that there is some level of redundancy
in sponsors for ports where hardware cannot be easily bought with Debian
money? That would cover the s390x port as well as any potential other
insanely-expensive-hardware port[1] that we might support now or in the
future.

If push comes to shove, we could also talk to IBM. Pretty much all POWER
hardware we have was sponsored by IBM; it's not unreasonable to assume
they might be willing to also sponsor s390x time or hardware.

[1] not that I know of any, but hey, you never know.

-- 
< ron> I mean, the main *practical* problem with C++, is there's like a dozen
   people in the world who think they really understand all of its rules,
   and pretty much all of them are just lying to themselves too.
 -- #debian-devel, OFTC, 2016-02-12