Date: Fri, 7 Sep 2018 11:52:24 -0400 From: Mark Johnston <markj@freebsd.org> To: Alexander Leidinger <Alexander@leidinger.net> Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r338507 - in head/sys: sys vm Message-ID: <20180907155224.GA63224@raichu> In-Reply-To: <20180907095010.Horde.LHANrMIgb7RjrRNYbtaBqYT@webmail.leidinger.net> References: <201809061928.w86JSrSb067799@repo.freebsd.org> <20180907095010.Horde.LHANrMIgb7RjrRNYbtaBqYT@webmail.leidinger.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Sep 07, 2018 at 09:50:10AM +0200, Alexander Leidinger wrote: > Quoting Mark Johnston <markj@freebsd.org> (from Thu, 6 Sep 2018 > 19:28:53 +0000 (UTC)): > > > Author: markj > > Date: Thu Sep 6 19:28:52 2018 > > New Revision: 338507 > > URL: https://svnweb.freebsd.org/changeset/base/338507 > > > > Log: > > Avoid resource deadlocks when one domain has exhausted its memory. Attempt > > other allowed domains if the requested domain is below the minimum paging > > threshold. Block in fork only if all domains available to the forking > > thread are below the severe threshold rather than any. > > Could the problem you fixed manifest itself in e.g. a hanging system > due to not enough memory free despite top displaying a lot of memory > free (let's assume 2 memory domains and several gigs in both)? Yes. > I have a system which never went below a specific amount of memory > (before the fix of the no-rebalancing bug in the scheduler) if NUMA > was enabled. With NUMA disabled I see much less memory free. I haven't > tried NUMA after the fix of the scheduler, but would give it a try > again if you think this could be a manifestation of the problem. I think it's worth trying again.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180907155224.GA63224>