spurious out-of-swap messages (plenty of swap available)

[tech-lists]

Hi,

A desktop system running latest 12-stable and poudriere-devel gives "out of
swap" errors when compiling some ports like rustc. But swap doesn't go over
7%, there's 32GB RAM and over 64GB of swap (2x 32GB swapfiles on an SSD plus
one 4GB swap partition, also on SSD):

% sysctl -a | grep swap
2 PART ada0s1b 3834175488 512 i 2 o 236223201280 ty freebsd-swap xs BSD xt 1
0 MD md98 34359738368 512 u 98 s 512 f 0 fs 0 l 34359738368 t vnode file
/swap1 label 
0 MD md99 34359738368 512 u 99 s 512 f 0 fs 0 l 34359738368 t vnode file
/swap0 label 
z0xfffff8001e3ae600 [shape=box,label="SWAP\nswap\nr#4"];
      <file>/swap1</file>
      <file>/swap0</file>
      <name>swap</name>
      <type>freebsd-swap</type>
vm.swap_enabled: 1
vm.domain.0.stats.unswappable: 0
vm.swap_idle_threshold2: 10
vm.swap_idle_threshold1: 2
vm.swap_idle_enabled: 0
vm.disable_swapspace_pageouts: 0
vm.stats.vm.v_swappgsout: 3034133
vm.stats.vm.v_swappgsin: 462625
vm.stats.vm.v_swapout: 428651
vm.stats.vm.v_swapin: 143721
vm.nswapdev: 3
vm.swap_fragmentation: 
vm.swap_async_max: 4
vm.swap_maxpages: 65126576
vm.swap_total: 72553652224
vm.swap_reserved: 76193538048

I see this in /var/log/messages:

Dec 19 15:38:07 desktop kernel: pid 57027 (rustc), jid 13, uid 0, was killed:
out of swap space
Dec 19 15:39:47 desktop kernel: pid 35795 (rustc), jid 13, uid 0, was killed:
out of swap space

Why isn't all the swap space being used? I could understand if it went to like
95% use then complained, but I've kept a close eye on it while compiling and
it doesn't get anywhere near that. As far as I'm aware, sysctl tunables are
default. I'm thinking maybe the out-of-swapspace messages mask something else?

thanks,
-- 
J.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-questions/attachments/20191219/d1482849/attachment.sig>