Re: A tale of systemd and MaxProcs

2016-10-04 Thread Denys Vlasenko
On 09/24/2016 08:34 PM, Zbigniew Jędrzejewski-Szmek wrote: On Thu, Sep 22, 2016 at 12:18:53PM -0400, Matthew Miller wrote: On Wed, Sep 21, 2016 at 12:32:40PM -0600, Kevin Fenzi wrote: * IMHO the initial upstream default didn't make sense for Fedora On this specific change, I'm not sure the *u

Re: A tale of systemd and MaxProcs

2016-09-26 Thread Matthew Miller
On Mon, Sep 26, 2016 at 03:53:53PM +, Zbigniew Jędrzejewski-Szmek wrote: > Ack. I run a lot of builds on rawhide, but none of my systems is large > enough to hit such issues. From my side I'll try to watch out for > changes in systemd and that change resource or privilege limitations > that cou

Re: A tale of systemd and MaxProcs

2016-09-26 Thread Zbigniew Jędrzejewski-Szmek
On Sun, Sep 25, 2016 at 01:11:16PM -0600, Kevin Fenzi wrote: > On Sun, 25 Sep 2016 16:54:45 - > "Timothy Ward" wrote: > > > Although I agree with the above then may be a builder should have a > > rawhide VM running with these updated critical packages if only for > > testing purposes to expos

Re: A tale of systemd and MaxProcs

2016-09-26 Thread Matthew Miller
On Sat, Sep 24, 2016 at 06:34:36PM +, Zbigniew Jędrzejewski-Szmek wrote: > > far-reaching infrastructure packages (systemd, glibc, kernel, whatever) > > that any new restrictions or constraints should be disabled by default > > in Fedora, regardless of upstream defaults, until we're able to hav

Re: A tale of systemd and MaxProcs

2016-09-25 Thread Kevin Fenzi
On Sun, 25 Sep 2016 16:54:45 - "Timothy Ward" wrote: > Although I agree with the above then may be a builder should have a > rawhide VM running with these updated critical packages if only for > testing purposes to expose this type of failure. Sure, as I noted perhaps we should move some of

Re: A tale of systemd and MaxProcs

2016-09-25 Thread Timothy Ward
> On Sat, Sep 24, 2016 at 2:34 PM, Zbigniew Jędrzejewski-Szmek > > That's a reflection on the nature of the change itself though. People > don't run large, stable workloads that would hit the TasksMax limit on > rawhide. They run them on stable releases. It's a chicken-egg > situation from a t

Re: A tale of systemd and MaxProcs

2016-09-25 Thread Kevin Fenzi
On Fri, 23 Sep 2016 07:18:47 -0700 Laura Abbott wrote: > The particular issue came here with an upgrade to F24. Where does > rawhide fit into this? It seems like this should have been found > with rawhide or at least someone noticed an issue. Well, I guess no one is running services that hit the

Re: A tale of systemd and MaxProcs

2016-09-25 Thread Josh Boyer
On Sat, Sep 24, 2016 at 2:34 PM, Zbigniew Jędrzejewski-Szmek wrote: > Nevertheless, for our defense, let me point out that the problems with > this change are obvious only in hindsight: after all new systemd versions > are only introduced in rawhide / pre-beta, and this was added in 228, > while F

Re: A tale of systemd and MaxProcs

2016-09-24 Thread Zbigniew Jędrzejewski-Szmek
On Thu, Sep 22, 2016 at 12:18:53PM -0400, Matthew Miller wrote: > On Wed, Sep 21, 2016 at 12:32:40PM -0600, Kevin Fenzi wrote: > > * IMHO the initial upstream default didn't make sense for Fedora > > On this specific change, I'm not sure the *updated* default makes sense > either. It still is quit

Re: A tale of systemd and MaxProcs

2016-09-23 Thread Laura Abbott
On 09/22/2016 09:18 AM, Matthew Miller wrote: On Wed, Sep 21, 2016 at 12:32:40PM -0600, Kevin Fenzi wrote: * IMHO the initial upstream default didn't make sense for Fedora On this specific change, I'm not sure the *updated* default makes sense either. It still is quite constrained. * Perhaps

Re: A tale of systemd and MaxProcs

2016-09-23 Thread Nico Kadel-Garcia
On Thu, Sep 22, 2016 at 12:18 PM, Matthew Miller wrote: > On Wed, Sep 21, 2016 at 12:32:40PM -0600, Kevin Fenzi wrote: >> * IMHO the initial upstream default didn't make sense for Fedora > > On this specific change, I'm not sure the *updated* default makes sense > either. It still is quite constra

Re: A tale of systemd and MaxProcs

2016-09-22 Thread Matthew Miller
On Wed, Sep 21, 2016 at 12:32:40PM -0600, Kevin Fenzi wrote: > * IMHO the initial upstream default didn't make sense for Fedora On this specific change, I'm not sure the *updated* default makes sense either. It still is quite constrained. > * Perhaps after beta but before final we ping maintainer

Re: A tale of systemd and MaxProcs

2016-09-21 Thread Florian Weimer
On 09/21/2016 08:32 PM, Kevin Fenzi wrote: Not this time. ;) The kernel maintainers started getting builds that failed and got stuck in odd ways. A few other larger packages (glibc) maintainers also hit this issue. The mock logs simply showed a bunch of "Fork failed" messages and a bunch of proce

A tale of systemd and MaxProcs

2016-09-21 Thread Kevin Fenzi
Greetings. I thought I would share this in hopes we can come up with ways to handle this sort of thing better. I do not blame anyone and I'd like to focus on positive steps we can take rather than pointing fingers. Just before the Fedora 25 development cycle started, I went and upgraded our koj