Re: Question about the GitLab runners at cl.net

Robert Goldman <[email protected]> Thu, 30 Apr 2026 10:10:15 -0500
Newsgroups gmane.lisp.common-lisp-net.devel
Message-ID <[email protected]>
--=_MailMate_B9821972-0FC2-4F76-9F50-AB9A395C8E96_=
Content-Type: text/plain; charset=UTF-8; format=flowed; markup=markdown
Content-Transfer-Encoding: quoted-printable

I have collected data about the comparison between the runs on =

common-lisp.net and sift.net.

Because this information is bulky, I have put it in a wiki page here:
https://gitlab.common-lisp.net/asdf/asdf/-/wikis/ASDF-test-failures-on-co=
mmon-lisp.net

Please have a look when time permits.  This shows the big divergence =

between cl.net -- failing after 1h -- and sift -- completing =

successfully in less than 6 minutes.  I can't account for it, but I have =

found some lines in the transcript that are separated by substantial =

amounts of time.

The CMU CL problems involve something different. I will upload log files =

for this, too, asap.  It takes a while to wrestle with the log files and =

markdown.

On 28 Apr 2026, at 18:14, David Cooper wrote:

> Hi Robert,=C2=A0
>
>
>
>
> Any "insight" i would claim would be speculation.=C2=A0
>
>
>
> Can you watch the log vs. the one you run at home in under 3 minutes? =

> Might the log reveal one or more bottlenecks in the build or test =

> process? (this is standard ASDF test suite you're talking about?)=C2=A0=

>
> =C2=A02+ minutes vs. timing out up to an hour points to something =

> pathological going on, our CPUs can't be _that_ much slower.=C2=A0
>
>
>
>
> Dave Cooper
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> From: Robert Goldman <[email protected]>
> To: "Discussion about the development of common-lisp.net,   the site =

> and the project."<[email protected]>
> Date: Tue, 28 Apr 2026 18:02:09 -0400
> Subject: Question about the GitLab runners at cl.net
>
>
>
> I have been having troubles with the ASDF tests on GitLab.  The ASDF =

> tests are timing out reliably on Clozure Common Lisp after an hour.  =

> The same tests on a local Linux server takes only 2m30s.
>
> This machine has 16=C2=A0=C2=A0=C2=A0=C2=A0cores, but the ASDF tests ar=
e not =

> multi-threaded.  The processors are Intel(R) Xeon(R) CPU E5-2680 v3 @ =

> 2.50GHz
>
> Those are pretty good, but by no means cutting edge.
>
> So I'm wondering if there's anything funny about the configuration of =

> the GitLab runner.
>
> Thanks for any insight.
>
> If I can, I will mirror this repo locally and see how the runtime =

> differs on our runner.

--=_MailMate_B9821972-0FC2-4F76-9F50-AB9A395C8E96_=
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE html>
<html>
<head>
<meta http-equiv=3D"Content-Type" content=3D"text/xhtml; charset=3Dutf-8"=
>
</head>
<body><div style=3D"font-family: sans-serif;"><div class=3D"markdown" sty=
le=3D"white-space: normal;">
<p dir=3D"auto">I have collected data about the comparison between the ru=
ns on common-lisp.net and sift.net.</p>
<p dir=3D"auto">Because this information is bulky, I have put it in a wik=
i page here:<br>
<a href=3D"https://gitlab.common-lisp.net/asdf/asdf/-/wikis/ASDF-test-fai=
lures-on-common-lisp.net" style=3D"color: #3983C4;">https://gitlab.common=
-lisp.net/asdf/asdf/-/wikis/ASDF-test-failures-on-common-lisp.net</a></p>=

<p dir=3D"auto">Please have a look when time permits.  This shows the big=
 divergence between cl.net -- failing after 1h -- and sift -- completing =
successfully in less than 6 minutes.  I can't account for it, but I have =
found some lines in the transcript that are separated by substantial amou=
nts of time.</p>
<p dir=3D"auto">The CMU CL problems involve something different. I will u=
pload log files for this, too, asap.  It takes a while to wrestle with th=
e log files and markdown.</p>
<p dir=3D"auto">On 28 Apr 2026, at 18:14, David Cooper wrote:</p>
<blockquote style=3D"margin: 0 0 5px; padding-left: 5px; border-left: 2px=
 solid #777777; color: #777777;">
<p dir=3D"auto">Hi Robert,=C2=A0</p>
<p dir=3D"auto">Any &quot;insight&quot; i would claim would be speculatio=
n.=C2=A0</p>
<p dir=3D"auto">Can you watch the log vs. the one you run at home in unde=
r 3 minutes? Might the log reveal one or more bottlenecks in the build or=
 test process? (this is standard ASDF test suite you're talking about?)=C2=
=A0</p>
<p dir=3D"auto">=C2=A02+ minutes vs. timing out up to an hour points to s=
omething pathological going on, our CPUs can't be <em>that</em> much slow=
er.=C2=A0</p>
<p dir=3D"auto">Dave Cooper</p>
<p dir=3D"auto">From: Robert Goldman <a href=3D"mailto:[email protected]=
o" style=3D"color: #777777;">[email protected]</a><br>
To: &quot;Discussion about the development of common-lisp.net,   the site=
 and the project.&quot;<a href=3D"mailto:[email protected]" style=
=3D"color: #777777;">[email protected]</a><br>
Date: Tue, 28 Apr 2026 18:02:09 -0400<br>
Subject: Question about the GitLab runners at cl.net</p>
<p dir=3D"auto">I have been having troubles with the ASDF tests on GitLab=
=2E  The ASDF tests are timing out reliably on Clozure Common Lisp after =
an hour.  The same tests on a local Linux server takes only 2m30s.</p>
<p dir=3D"auto">This machine has 16=C2=A0=C2=A0=C2=A0=C2=A0cores, but the=
 ASDF tests are not multi-threaded.  The processors are Intel(R) Xeon(R) =
CPU E5-2680 v3 @ 2.50GHz</p>
<p dir=3D"auto">Those are pretty good, but by no means cutting edge.</p>
<p dir=3D"auto">So I'm wondering if there's anything funny about the conf=
iguration of the GitLab runner.</p>
<p dir=3D"auto">Thanks for any insight.</p>
<p dir=3D"auto">If I can, I will mirror this repo locally and see how the=
 runtime differs on our runner.</p>
</blockquote>

</div>
</div>
</body>

</html>

--=_MailMate_B9821972-0FC2-4F76-9F50-AB9A395C8E96_=--