Date: Mon, 18 Jun 2018 23:55:42 -0700 From: bob prohaska <fbsd@www.zefox.net> To: Jukka Ukkonen <jau789@gmail.com> Cc: Mark Millard <marklmi@yahoo.com>, freebsd-arm@freebsd.org, bob prohaska <fbsd@www.zefox.net> Subject: Re: GPT vs MBR for swap devices Message-ID: <20180619065542.GA85994@www.zefox.net> In-Reply-To: <0C802675-9DE2-4446-B0F1-528D40C69C68@gmail.com> References: <7AB401DF-7AE4-409B-8263-719FD3D889E5@yahoo.com> <20180618230419.GA81275@www.zefox.net> <A8D00616-ADA7-4A33-8787-637AFEF547CF@yahoo.com> <20180619005519.GB81275@www.zefox.net> <BC42DDF9-9383-437B-8AE2-A538050C5160@yahoo.com> <20180619034232.GA81800@www.zefox.net> <0C802675-9DE2-4446-B0F1-528D40C69C68@gmail.com>
index | next in thread | previous in thread | raw e-mail
On Tue, Jun 19, 2018 at 07:06:18AM +0300, Jukka Ukkonen wrote: > > Are you sure it is not /usr/obj activity which you are seeing when > there are large write delays? I am absolutely not sure. > On systems using traditional spinning disks for everything else > it really makes sense to put /usr/obj on its own SSD making sure > the SSD does not share an I/O channel with any other device. > Agreed that spreading throughput among channels is good. That's why I tried to split swap between microSD and USB. The RPI2 systems I've set up before use a USB flash drive for /usr, /var, /tmp and swap and they build world successfully. When I set up the RPI3 I thought it would be an improvement to split the swap between partitions on microSD and the USB flash drive. The suspicion I'm harboring is that the swapper on the Pi3 is somehow confounded by having two separate swap devices. The simplest test I can easily perform is to use one swap partiton on the microSD card and a second swap partition on a separate USB device. To start with that'll be a mechanical hard disk, which is generally accepted as well-behaved. Thanks for reading! bob prohaskahelp
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180619065542.GA85994>