[OpenAFS-devel] qemu arm slave for buildbot
Phillip Moore
w.phillip.moore@gmail.com
Tue, 2 Nov 2010 09:12:57 -0400
--20cf3005dc5cf0fe77049411ae34
Content-Type: text/plain; charset=ISO-8859-1
I'm curious why the i386 Linux distros are considered Tier 1, yet you don't
say anything about x86_64. Coming from the world of financial computing,
where we run just about everything on x86_64 hardware, I'm surprised you
only mention i386.
Seems to me that what is most important is really the OS distro in this case
(Debian, RHEL) and both i386 and x86_64 versions should be tested as Tier 1
platforms.
On Tue, Nov 2, 2010 at 8:57 AM, Simon Wilkinson <sxw@inf.ed.ac.uk> wrote:
>
> On 2 Nov 2010, at 01:29, Jason Edgecombe wrote:
>
> > hi everyone,
> >
> > I'm working with someone who is offering a qemu arm buildslave. The
> build on the qemu arm hosts takes ~3 hours. I'm hesitant to have all of the
> gerrit submissions take three hours to build. A daily build of the git
> branches might be better in this case.
>
> Yes.
>
> I'm already finding the cycle time of the IRIX build slave hugely
> frustrating, and would be opposed to adding anything that lengthens that
> time further.
>
> I think build cycles of longer than about 20 minutes are going to be
> unsustainable in the long term - we should aim to put anything that takes
> longer than that into their own build queues that just build the various
> branch tips daily. This is similar to what other folks do with Tier 1 and
> Tier 2 platforms.
>
> So, I guess that I'm proposing that we have i386 Debian, i386 RHEL, Darwin,
> Windows and Solaris as Tier 1 systems - we aim to have buildbot coverage for
> all of these, and to only push changes that buildbot verifies across all of
> those platforms.
>
> We then have Tier 2 systems - Irix, your new Arm build slave. We accept
> that master will break on these systems, but will endeavour to pick up the
> pieces after buildbot has notified them.
>
> And finally, we have Tier 3 systems - ones for which we have no buildbot
> coverage, and where we only pick up breakage when we're notified of it,
> either by users, or during the pre-release smoke testing.
>
> Comments?
>
> Simon.
>
>
>
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel
>
--20cf3005dc5cf0fe77049411ae34
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: quoted-printable
<div><br></div>I'm curious why the i386 Linux distros are considered Ti=
er 1, yet you don't say anything about x86_64. =A0 Coming from the worl=
d of financial computing, where we run just about everything on x86_64 hard=
ware, I'm surprised you only mention i386. =A0<div>
<br></div><div>Seems to me that what is most important is really the OS dis=
tro in this case (Debian, RHEL) and both i386 and x86_64 versions should be=
tested as Tier 1 platforms. =A0=A0<br><br><div class=3D"gmail_quote">On Tu=
e, Nov 2, 2010 at 8:57 AM, Simon Wilkinson <span dir=3D"ltr"><<a href=3D=
"mailto:sxw@inf.ed.ac.uk">sxw@inf.ed.ac.uk</a>></span> wrote:<br>
<blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1p=
x #ccc solid;padding-left:1ex;"><div class=3D"im"><br>
On 2 Nov 2010, at 01:29, Jason Edgecombe wrote:<br>
<br>
> hi everyone,<br>
><br>
> I'm working with someone =A0who is offering a qemu arm buildslave.=
The build on the qemu arm hosts takes ~3 hours. I'm hesitant to have a=
ll of the gerrit submissions take three hours to build. A daily build of th=
e git branches might be better in this case.<br>
<br>
</div>Yes.<br>
<br>
I'm already finding the cycle time of the IRIX build slave hugely frust=
rating, and would be opposed to adding anything that lengthens that time fu=
rther.<br>
<br>
I think build cycles of longer than about 20 minutes are going to be unsust=
ainable in the long term - we should aim to put anything that takes longer =
than that into their own build queues that just build the various branch ti=
ps daily. This is similar to what other folks do with Tier 1 and Tier 2 pla=
tforms.<br>
<br>
So, I guess that I'm proposing that we have i386 Debian, i386 RHEL, Dar=
win, Windows and Solaris as Tier 1 systems - we aim to have buildbot covera=
ge for all of these, and to only push changes that buildbot verifies across=
all of those platforms.<br>
<br>
We then have Tier 2 systems - Irix, your new Arm build slave. We accept tha=
t master will break on these systems, but will endeavour to pick up the pie=
ces after buildbot has notified them.<br>
<br>
And finally, we have Tier 3 systems - ones for which we have no buildbot co=
verage, and where we only pick up breakage when we're notified of it, e=
ither by users, or during the pre-release smoke testing.<br>
<br>
Comments?<br>
<font color=3D"#888888"><br>
Simon.<br>
</font><div><div></div><div class=3D"h5"><br>
<br>
<br>
_______________________________________________<br>
OpenAFS-devel mailing list<br>
<a href=3D"mailto:OpenAFS-devel@openafs.org">OpenAFS-devel@openafs.org</a><=
br>
<a href=3D"https://lists.openafs.org/mailman/listinfo/openafs-devel" target=
=3D"_blank">https://lists.openafs.org/mailman/listinfo/openafs-devel</a><br=
>
</div></div></blockquote></div><br></div>
--20cf3005dc5cf0fe77049411ae34--