On Oct 24 10:01:42, h...@stare.cz wrote:
> > Does the swap partition sd0b actually exist,
> > or are you by any chance running these machines swapless?
>
> Indeed, I am running without swap.
>
> > I've seen this seemingly random unstable behavior
> > on our RockPI 4a SBCs when no swap is configured.
>
> I can confirm that after adding a 1GB swap partition,
> these random killings have completely disappeared.
> Thank you! (I will also try on the RPI4.)
Yep: adding 256MB of sd0b to the RPI4 makes these disappear
- just did a kernel build. Thanks again.
Jan
> > Just to be clear - the lack of a swap partition
> > causes this behaviour even when plenty of physical RAM is available.
>
> Yes, that seems to be my case too.
> I don't understand it though:
>
> $ swapctl -l
> Device 512-blocks Used Avail Capacity Priority
> /dev/sd0b 2104516 0 2104516 0% 0
>
> $ swapctl -s
> total: 2104516 512-blocks allocated, 0 used, 2104516 available
>
> The swap never gets used. For example, I just did full
> kernel build, and swapctl -s still says zero.
>
> If there is enough RAM to not have to swap,
> why would a missing swap get processes killed
> (if that was actualy the root cause of course)?
>
> > In our case, creating a swap parition of a mere 32 MB
> > mitigated the issue, even though it wasn't touched doing
> > a kernel compile with 4 Gb of physical RAM.
>
> I will try with a very small swap partition just to confirm;
> in fact, the installer has offered to create a ~32MB swap
> on a 2GB card before and I foolishly did not use it,
> thinking I will not need swap anyway.
>
> Is this known? Or is there another reason the installer
> wants to create e.g. a 32 MB swap on a machine with 1GB RAM
> and a few GB of disk?
>
> There is something else going on on the swap partition,
> even of not used for lack of physical ram: for example,
> now that I run with the swap partition, I can remove it
> with swapctl -d, but cannot remove it from the disklabel
> because disklabel: DIOCWDINFO: Device busy.
