Date: Sat, 30 Jun 2018 19:53:54 -0700 From: bob prohaska <fbsd@www.zefox.net> To: Trev <freebsd-arm@sentry.org> Cc: freebsd-arm@freebsd.org, bob prohaska <fbsd@www.zefox.net> Subject: Re: RPI3 swap experiments Message-ID: <20180701025354.GA49303@www.zefox.net> In-Reply-To: <f842b145-c79b-161d-5282-992212f8296f@sentry.org> References: <20180627054027.GA22144@www.zefox.net> <CC72E766-03CB-476C-8F2B-DAAC266CE63D@yahoo.com> <20180627194217.GA27793@www.zefox.net> <C6303FC5-B412-472C-98E4-9A1E45C38535@yahoo.com> <20180628022457.GA30110@www.zefox.net> <7B9D272D-3EDE-46FA-8A1C-AEE65047167C@yahoo.com> <20180628163328.GA33408@www.zefox.net> <51e208b4-9f14-58f7-1e70-6ef8db2c0bed@sentry.org> <20180629155131.GA35717@www.zefox.net> <f842b145-c79b-161d-5282-992212f8296f@sentry.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Jul 01, 2018 at 09:42:15AM +1000, Trev wrote: > bob prohaska wrote on 30/06/2018 01:51: > > On Fri, Jun 29, 2018 at 10:28:25PM +1000, Trev wrote: > >> > >> Jun 29 20:20:35 rpi3 kernel: pid 66844 (c++), uid 0, was killed: out of > >> swap space > >> > >> At the time of "out of swap" only 157M of the 4G swap was in use. The > >> 4GB swap partition is on a 4GB USB2 memory key (da1p1), everything else > >> is on the sdcard. There are no excessively long ms/w being recorded. > >> There are no USB or other related errors. > >> > > > > That's a little surprising, and disturbing. By my experience your setup > > should have run to completion. > > > > Do you see a "too much swap" complaint? I don't know if it's a problem, > > but 4 GB of swap triggers one on my RPi3. 3 GB does not. 2 GB is enough > > to finish -j4 buildworld. > > Jun 30 23:32:12 rpi3 kernel: pid 60735 (c++), uid 0, was killed: out of > swap space > > No go with 2GB (I deleted the da0p1 swap partition; destroyed da0; then > remade it it with a 2GB p1; then rebooted - no "too much swap" > complaints). Out of swap with 56M of 2GB used. And this was with make > -j2 buildworld. > It just crossed my mind that the successful -j4 buildworlds that used USB swap (both mechanical and flash) separate from /usr and /var placed the swap on a powered USB hub. It would be exceeding strange if that proved helpful, but as things are going.... The other successful -j4 buildworlds put swap on the root microSD card. I'd rate that a harder experiment, even if there is free space on the card. My talents with gpart leave much to be desired. If you're properly gifted it might also be worth trying. > Here's the top and gstat stats leading up to, at the time of out of swap > and just after: > > last pid: 61290; load averages: 2.42, 2.27, 2.19 up 0+12:23:41 > 23:32:07 > 33 processes: 2 running, 31 sleeping > > Mem: 609M Active, 296K Inact, 95M Laundry, 176M Wired, 97M Buf, 20M Free > Swap: 2048M Total, 47M Used, 2001M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 4 68 0 0 0.0 68 1487 20.9 36.7 da0p1 > > last pid: 61294; load averages: 2.42, 2.27, 2.19 up 0+12:23:42 > 23:32:08 > 33 processes: 2 running, 31 sleeping > > Mem: 612M Active, 180K Inact, 94M Laundry, 176M Wired, 97M Buf, 19M Free > Swap: 2048M Total, 49M Used, 1999M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 4 110 0 0 0.0 110 3653 24.3 77.8 da0p1 > > last pid: 61298; load averages: 2.42, 2.27, 2.19 up 0+12:23:43 > 23:32:09 > 33 processes: 2 running, 31 sleeping > > Mem: 614M Active, 248K Inact, 90M Laundry, 176M Wired, 97M Buf, 20M Free > Swap: 2048M Total, 53M Used, 1995M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 0 28 0 0 0.0 28 2737 157.7 113.7 da0p1 > > last pid: 61302; load averages: 2.42, 2.27, 2.19 up 0+12:23:45 > 23:32:11 > 33 processes: 2 running, 31 sleeping > > Mem: 322M Active, 297M Inact, 88M Laundry, 176M Wired, 97M Buf, 17M Free > Swap: 2048M Total, 55M Used, 1993M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 5 0 0 0 0.0 0 0 0.0 0.0 da0p1 > > last pid: 61306; load averages: 2.42, 2.27, 2.19 up 0+12:23:46 > 23:32:12 > 33 processes: 1 running, 32 sleeping > > Mem: 326M Active, 164K Inact, 379M Laundry, 176M Wired, 97M Buf, 19M Free > Swap: 2048M Total, 56M Used, 1992M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 6 0 0 0 0.0 0 0 0.0 0.0 da0p1 > > last pid: 61310; load averages: 2.39, 2.27, 2.19 up 0+12:23:47 > 23:32:13 > 33 processes: 1 running, 32 sleeping > > Mem: 331M Active, 184K Inact, 379M Laundry, 176M Wired, 97M Buf, 14M Free > Swap: 2048M Total, 56M Used, 1992M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 6 0 0 0 0.0 0 0 0.0 0.0 da0p1 > > [OUT OF SWAP at 23:32:14 captured immediately below] > > last pid: 61315; load averages: 2.39, 2.27, 2.19 up 0+12:23:48 > 23:32:14 > 33 processes: 1 running, 32 sleeping > > Mem: 78M Active, 176K Inact, 315M Laundry, 174M Wired, 97M Buf, 332M Free > Swap: 2048M Total, 56M Used, 1992M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 0 48 37 710 126.2 11 1186 341.9 26.4 da0p1 > > last pid: 61320; load averages: 2.39, 2.27, 2.19 up 0+12:23:49 > 23:32:15 > 33 processes: 2 running, 31 sleeping > > Mem: 55M Active, 848K Inact, 1644K Laundry, 174M Wired, 97M Buf, 669M Free > Swap: 2048M Total, 48M Used, 2000M Free, 2% Inuse > 0 0 0 0 0.0 0 0 0.0 0.0 mmcsd0s2a > 0 0 0 0 0.0 0 0 0.0 0.0 da0p1 The delays look much better than I've seen on successful -j4 buildworlds.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180701025354.GA49303>