Poudriere Timeout

[Bryan Drewery]

On 1/22/2015 6:34 PM, Cy Schubert wrote:
> In message <54C18193.1000701 at FreeBSD.org>, Bryan Drewery writes:
>> This is an OpenPGP/MIME signed message (RFC 4880 and 3156)
>> --NvIRPgp2pVcRmjhwtM9TwAvH1900GK90a
>> Content-Type: text/plain; charset=windows-1252
>> Content-Transfer-Encoding: quoted-printable
>>
>> On 1/18/2015 7:45 PM, Cy Schubert wrote:
>>> Hi,
>>> =20
>>> =20
>>> Has anyone seen this before?=20
>>> =20
>>> print/texlive-texmf texlive-texmf-20140525_4 package/timeout runaway_pr=
>> ocess
>>> =20
>>> =20
>>> /usr/bin/touch /wrkdirs/usr/ports/print/texlive-texmf/work/stage/usr/lo=
>> cal/s
>>> hare
>>> /texmf-dist/doc/.keep_me
>>> /bin/mkdir -p /wrkdirs/usr/ports/print/texlive-texmf/work/stage/usr/loc=
>> al/sh
>>> are/
>>> texmf-dist/source
>>> /usr/bin/touch /wrkdirs/usr/ports/print/texlive-texmf/work/stage/usr/lo=
>> cal/s
>>> hare
>>> /texmf-dist/source/.keep_me
>>> =3D=3D=3D=3D> Compressing man pages (compress-man)
>>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
>> =3D
>>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D<p=
>> hase: package        >=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>>> =3D=3D=3D>  Building package for texlive-texmf-20140525_4
>>> =3D=3D=3D=3D>> Killing timed out build after 3600 seconds
>>> =3D=3D=3D=3D>> Cleaning up wrkdir
>>> =3D=3D=3D>  Cleaning for texlive-texmf-20140525_4
>>> build of print/texlive-texmf ended at Fri Jan 16 23:49:09 PST 2015
>>> build time: 02:41:43
>>> !!! build failure encountered !!!
>>> =20
>>> Building the port by hand, make package, took just over 19 minutes (115=
>> 0=20
>>> seconds). Would increasing the timeout from 3600 to something larger (l=
>> ike=20
>>> 7200 -- I know this will require hacking the code) address the issue?
>>> =20
>>> =20
>>
>> I'm curious why building outside of poudriere took 19 minutes, but
>> inside took over 60 minutes. Did you have MFS or TMPFS enabled at the
>> time of 60+ minutes?
> 
> No. I don't have enough memory to support mfs or tmpfs except for /tmp. 
> Having said that, my poudriere jails don't have tmpfs /tmp while all my 
> servers (and laptop) do.
> 
>>
>> Is pkg the same version in the jail as your host? It may be a recent pkg
>> create regression.
> 
> All the software is the same. The difference is that the failed build was 
> on an AMD 64 X2 4600+ (dual core with 5.5 GB) whereas the successful build 
> was on an Intel Core I3 dual core with four threads with 6 GB. All 
> filesystems are ZFS.
> 
> My guess is that the speed of the processors may have had something to do 
> with it. I don't think the extra 500 MB ram mattered much. All my build 
> systems are amd64 architecture.
> 
> 

My guess is that MAKE_JOBS was disabled for the build inside of
Poudriere. By default (for bulk) it will parallelizing building multiple
packages at once, and each port build gets only -j1. This is something I
would like to improve with dynamic cpusets and/or racct type things. I
just haven't gotten there yet.

testport defaults to using MAKE_JOBS though.

-- 
Regards,
Bryan Drewery

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/freebsd-ports/attachments/20151114/1b250792/attachment.bin>