On Mon, Aug 14, 2023 at 02:00:49PM +1000, Paul Ripke wrote:
> I recently upgraded my netbsd-9 amd64 w 16GiB RAM to netbsd-10. Which, btw,
> feels much "snappier", anecdotally.
> After running a few builds and whatnot, I chased down where all my RAM was
> going. Manually sorted:
>
> Memory resource pool statistics
> Name Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg Maxpg
> Idle
> buf2k 2048 399771 0 97706 164994 11676 153318 164983 0 inf
> 106
> vcachepl 640 1412222 0 612635 183127 39581 143546 183127 0 inf
> 0
> ffsino 280 1440870 0 641796 78452 14088 64364 78452 0 inf
> 0
> ffsdino2 264 1445733 0 646660 73222 13218 60004 73222 0 inf
> 0
> namecache 192 1257514 0 444866 39817 2 39815 39817 0 inf
> 0
> buf16k 16896 177282 0 71409 35542 250 35292 35292 0 inf
> 1
> bufpl 280 425741 0 12096 30334 490 29844 30334 0 inf
> 0
> anonpl 40 4816174 0 3570206 24466 2 24464 24466 0 inf
> 0
> rwlock 64 1649569 0 753596 18923 2298 16625 18923 0 inf
> 0
> mutex 64 1337583 0 537463 17451 1218 16233 17451 0 inf
> 0
> dirhashblk 2048 30827 0 2014 15238 732 14506 15238 0 inf
> 64
> radixnode 192 393557 0 273821 10589 1303 9286 10589 0 inf
> 12
> pvpage 4096 14790 0 7333 10066 2131 7935 8964 0 inf
> 478
> kmem-01024 1088 39226 0 20445 9647 3150 6497 8117 0 inf
> 0
> vmmpepl 192 213489 0 122047 6626 1562 5064 5625 0 inf
> 10
> kmem-02048 2048 22881 0 12785 8394 3346 5048 6698 0 inf
> 0
> mclpl 2048 33126 0 24568 5514 1211 4303 4390 0 261333
> 24
> buf4k 4096 7619 0 4102 3884 366 3518 3838 0 inf
> 1
> phpool-64 64 215806 0 22397 3283 64 3219 3283 0 inf
> 4
> mbpl 520 76469 0 58519 3265 520 2745 2795 0 inf
> 2
> ...
> Totals 177503409 0 169479744 816744 149009 667735
>
> In use 4025485K, total allocated 4834772K; utilization 83.3%
>
> That seems a little excessive, to my eye? Is this "normal"? Having >1/4 of RAM
> apparently used mostly for file metadata, if I'm reading this correctly? I
> don't recall the usage under -9, unfortunately, but I also never had a reason
> to look.
>
> I do set 'kern.maxfiles=40960', and also tweak 'vm.{anon,file}{min,max}', but,
> afaict, nothing else relevant.
>
> fwiw, this is a single raid0 over a couple of wd, all ffsv2, no zfs.
>
> Actually, looking again... this is my only system with such a large
> buf2k... what commonly allocates from that pool?
Following up, I see buf2k allocated when I run 'git status' under netbsd/pkgsrc,
netbsd/src, etc repos. I also see growth during /etc/daily. The filesystem is
ffsv2 with log, on raidframe raid1 of wd[01]a.
--
Paul Ripke
"Great minds discuss ideas, average minds discuss events, small minds
discuss people."
-- Disputed: Often attributed to Eleanor Roosevelt. 1948.
