RPI3 swap experiments ["was killed: out of swap space" with: "v_free_count: 5439, v_inactive_count: 1"]

[bob prohaska]

On Thu, Aug 09, 2018 at 09:42:59AM -0600, Warner Losh wrote:
> On Thu, Aug 9, 2018 at 9:37 AM, bob prohaska <fbsd at www.zefox.net> wrote:
> 
> >
> > There's a kernel building now with
> > options         CAM_IOSCHED_DYNAMIC
> > in the config file. Is it still worth trying? Anything else to try?
> >
> 
> It won't be a cure-all, out of the box, I don't think. However, the read
> biasing code may help sneak a few 'reads' in between writes which may help
> keep away from the pathological behavior.... Or not, it's hard to say...
> I've not looked at swapping to super-crappy nand (I mean thumb drives) in
> as much detail as the drives we use for work.
> 

Alas, 'tis true. Combined with the patches so far it didn't seem to help
at all, with or without using vm.pageout_oom_seq to 120. The -j4 buildworld
ended with an OOMA kill using 12 and a panic and reboot at 120. 

A -j4 buildworld actually got farther using vm.pageout_oom_seq=120 with no
other changes.

Results are at 
http://www.zefox.net/~fbsd/rpi3/swaptests/r337226M/1gbsdflash_1gbusbflash/batchqueue/pageout120/slow_swap/cam_iosched_dynamic/

Any suggestions on what to try next would be welcome.

Thanks for reading,

bob prohaska