From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joram Schrijver Subject: Re: guix-daemon hitting systemd TasksMax Date: Mon, 18 Apr 2016 22:42:00 +0200 Message-ID: <1461012120.3265374.582525641.595D96A9@webmail.messagingengine.com> References: <1460677052.2416958.579259849.502AC954@webmail.messagingengine.com> <20160416012901.GB12389@jasmine> <1460812335.1088225.580600441.5A60AF4E@webmail.messagingengine.com> <20160416161438.GB9412@jasmine> <871t65f9zb.fsf@gnu.org> <20160418203345.GB26832@jasmine> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: quoted-printable Return-path: Received: from eggs.gnu.org ([2001:4830:134:3::10]:35980) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1asFzh-0003Vv-II for guix-devel@gnu.org; Mon, 18 Apr 2016 16:42:06 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1asFzf-0004q3-1W for guix-devel@gnu.org; Mon, 18 Apr 2016 16:42:05 -0400 Received: from out2-smtp.messagingengine.com ([66.111.4.26]:54341) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1asFze-0004nB-U3 for guix-devel@gnu.org; Mon, 18 Apr 2016 16:42:02 -0400 Received: from compute1.internal (compute1.nyi.internal [10.202.2.41]) by mailout.nyi.internal (Postfix) with ESMTP id 8853820C26 for ; Mon, 18 Apr 2016 16:42:00 -0400 (EDT) In-Reply-To: <20160418203345.GB26832@jasmine> List-Id: "Development of GNU Guix and the GNU System distribution." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: guix-devel-bounces+gcggd-guix-devel=m.gmane.org@gnu.org Sender: "Guix-devel" To: Leo Famulari , =?ISO-8859-1?Q?Ludovic=20Court=E8s?= Cc: guix-devel@gnu.org Heh, no I'm not. I have to do it with 4 cores + hyperthreading. My first `guix pull` did have to build many things though, spawning numerous threads in some cases. I eye-sampled the thread counts during one try and they were certainly in the hundreds. I installed Guix from the AUR[1], so I was running the most recent release. Substitutes were enabled, but almost everything had to be built instead of downloaded. [1]: https://aur.archlinux.org/packages/guix/ -- Joram On Mon, Apr 18, 2016, at 10:33 PM, Leo Famulari wrote: > On Sun, Apr 17, 2016 at 12:46:16AM +0200, Ludovic Court=C3=A8s wrote: > > Leo Famulari skribis: > >=20 > > > On Sat, Apr 16, 2016 at 03:12:15PM +0200, Joram Schrijver wrote: > > >> I am running Arch Linux, also with systemd 229. It doesn't look like > > >> Debian applies any patches to systemd, so our installations should > > >> pretty much be the same. > > >>=20 > > >> Perhaps the limit is only reached during the build process of some > > >> specific packages? > > > > > > `guix pull` usually doesn't build any packages, but sometimes it does > > > need to. My (limited) understanding is that it sometimes needs to bui= ld > > > (or download) updated dependencies of Guix itself. > > > > > > Do you think the number of "Tasks" would increase with more cores? I > > > only have 4 cores on my machines. Maybe if you had more, the builders > > > spawned by your guix-daemon would run more tasks as a consequence of > > > using all the cores. > >=20 > > Right, =E2=80=98guix pull=E2=80=99 entails a build of Guix using as man= y cores as > > possible, with one process per core (and I think each Guile has two > > threads, one of which is the signal delivery thread.) > >=20 > > So, if the limit on the total number of threads + processes is 512 and > > you have 256 cores or more (lucky you!), you may hit the limit. >=20 > Joram, are you this lucky? ;)