Date: Sat, 17 Feb 2018 20:39:45 -0800 From: bob prohaska <fbsd@www.zefox.net> To: tech-lists <tech-lists@zyxst.net> Cc: freebsd-arm@freebsd.org, bob prohaska <fbsd@www.zefox.net> Subject: Re: Pi3 out of swap at < 50% Message-ID: <20180218043945.GA95271@www.zefox.net> In-Reply-To: <3d2bd5e3-4567-cdfe-04d4-b572371a212e@zyxst.net> References: <20180217162732.GA93736@www.zefox.net> <1518885801.91697.2.camel@freebsd.org> <9bf0ca6e-1916-3f4d-71fe-0515fe6a717f@zyxst.net> <20180218003840.GD93736@www.zefox.net> <3d2bd5e3-4567-cdfe-04d4-b572371a212e@zyxst.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Feb 18, 2018 at 03:38:49AM +0000, tech-lists wrote: > On 18/02/2018 00:38, bob prohaska wrote: > > The arm64 RPI3 also has 2 GB configured in two equal partitions, one > > on the USB flash drive and one on the microSD card used for booting. > > It's claiming "out of swap" with only about 50% usage. Swapinfo shows > > both partitions are being used, so that's not the problem. Kernel and > > world are a couple weeks out of sync, which might be a problem but it > > can't be checked until buildworld runs to completion, which it hasn't. > > In that situation I'd deactivate the swap on the card, and make a few > swap partitions on the flash drive. I used to run into the sort of > difficulty you're seeing if swap was on the microsd. It's not just swap > though; enough i/o activity on the card and it starts blocking. You can > tell it's happening if in another terminal if you type sync. It takes > several seconds (sometimes tens of seconds) to return. > Sounds as if I've been too clever by half 8-) The idea of splitting swap between usb flash and microSD was analogous to splitting swap across multiple hard drives, to improve average response speed. Your experience suggests I've made matters worse, not better, by congesting the traffic to the microSD. Still, watching gstat indicates that for the most part write times are under 10 ms and even when much larger the system doesn't report it's out of swap; those errors, at least on the RPi3, don't seem to correlate with much of anything if I can believe the free swap indicator in top. Probably the next thing I should try is simply disabling the microSD swap partition to see if system behavior makes more sense. > I'd like to be able to set async on the connected HD but it seems even > when it's set in fstab it doesn't show up when the mount command is > issued. The HD connects via a powered hub btw. > > These days I don't do updates or compiling on the rpi3 itself now that > poudriere/synth can cross-compile ports. If I need to update the OS I'll > use crochet to build an image from whatever sources on a beefy amd64 > then restore my data. I must have gone through dozens of microsd cards > before arriving at this solution. Some just completely fail, others > suddenly become read-only. > > There's probably a solution in dd'ing the card image to a more powerful > computer, mounting the image and crossbuilding into it which would be > more efficient I think but I haven't tried it yet. > -- My prime objective was to retire high power, high cost computers. So far the RPi2 computers running stable/11 have been quite good. The Pi3 running arm64 was tried because Pi2's compatible with armv7 seem no longer available and I didn't want to get stranded with no spares. Arm64 is a bigger leap than expected. Thanks for writing! bob prohaska
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180218043945.GA95271>