Date: Tue, 6 Mar 2018 11:28:17 -0800 (PST) From: "Rodney W. Grimes" <freebsd-rwg@pdx.rh.CN85.dnsmgr.net> To: bob prohaska <fbsd@www.zefox.net> Cc: "freebsd-arm@freebsd.org" <freebsd-arm@freebsd.org>, Ian Lepore <ian@freebsd.org> Subject: Re: Is maximum swap usage tunable? Message-ID: <201803061928.w26JSHmQ050295@pdx.rh.CN85.dnsmgr.net> In-Reply-To: <20180306190135.GB51223@www.zefox.net>
next in thread | previous in thread | raw e-mail | index | archive | help
> On Tue, Mar 06, 2018 at 10:03:21AM -0800, Rodney W. Grimes wrote: > > > > What are you using to measure the power with? > > The average DMM is not going to see the rapid pulses of > > noise that increase with current consumption that can > > lead to "odd behavior". You need a good and fairly > > fast scope to truely rule this out. > > Alas, no scope, just a DVM. But, there are four other RPis on > the same outlet strip, three running 11-stable are trouble free, > the fourth only started running out of swap with the latest revision > of -current.. THey are running this same work load too? It does not sound like they are all running the same code? Three on 11-stable, one on ? > > Also the GPIO header is a good electrical distance > > from the SOC and much of the noise would be filtered > > out before then. > That would qualify as a hardware failure, but what could > cause it, and why only after recent upgrades to the sources? Marginal hardware can have very slight tickle points, especially when it comes to noise on power, or an ever so slightly slow component. At 1Ghz+ frequencies it does not take much of a slight power supply noise change to go from a working siutaion to a not working. A software change can cause a change in the amount of nose generated. > > But atleast you have confirmed your not sagging VCC down > > to 4.75 or 4.5 as I have seen on some of the power power > > bricks used to run *Pi*'s with. > > > > It would probably be benificial if someone could duplicate > > to a reasonable degree your failure mode, this would make > > a hardware related problem less likely as the cause. > > > > I think that might be coming, what with Mike's lastest report in > this thread. Could the new thread Re: Strange ARC/Swap/CPU be > related? I haven't been watching wired memory, on the Pi3 it's > at 160MB about ten hours into buildworld. Idle time is zero, > swap usage is at 51MB with essentially no activity. If there are > other things to check please tell me. Are there any explict swap > performance tests that react quicker than make buildworld? Watch the output of the command I posted in the other thread, vmstat -z | egrep 'ITEM|g_bio|UMA' There was a known g_bio leak. -- Rod Grimes rgrimes@freebsd.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201803061928.w26JSHmQ050295>