Re: wired memory leak?

[P Gman <patrickgunsolley%gmail.com@localhost>]

Just for the tally, yes I've noticed similar patterns on a Ryzen 2400G
and an AMD A6. The usage appeared to be relative to yours given my
system's memory (16GB on the Ryzen and 8GB on the A6). Unfortunately I
don't have NetBSD installed on anything at the moment and cannot
provide actual numbers.

On Thu, Oct 29, 2026 at 2:57 PM Greg Troxel <gdt%lexort.com@localhost> wrote:
>
> I have a pretty normal system, 9th generation i7 with 32 GB of RAM.  The
> only semi-odd thing is that I run ZFS.  Less odd, the nextcloud sync
> client and syncthing.  It was crashy on NetBSD 10, but was solid on
> NetBSD 11 as soon as I updated.  (11 has fixes for flushing dirty mmap'd
> pages and some other locking issues, so this is not surprising.) I've
> been updating every month or so along NetBSD 11.
>
> I was rebuilding NetBSD on various branches, one at a time with -j2
> (should be not really noticeable).  I noticed firefox was very laggy,
> and then I noticed in top that the system was io bound and paging hard.
> Wired memory had begun to consume 75% of RAM.
>
>   Memory: 307M Act, 121M Inact, 24G Wired, 52M Exec, 306M File, 343M Free
>   Swap: 40G Total, 660M Used, 39G Free / Pools: 6949M Used / Network: 4K In, 3K Out
>
> After rebooting, it was "only" a few hundred MB.
>
> Just now, after not particularly  hard use
>
>   5:43PM  up 1 day,  8:13, 27 users, load averages: 0.38, 0.65, 0.52
>
>   Memory: 12G Act, 777M Inact, 2070M Wired, 523M Exec, 9301M File, 3701M Free
>   Swap: 40G Total, 40G Free / Pools: 12G Used / Network: 10K In, 6K Out
>
> which is not trouble yet but heading in the wrong direction.  I had not
> started syncthing, trying to simplify.
>
> I was the first time, I think, running kernel from September 29, and now
> am on sources from October 28 about 12Z.
>
> I do not remeber this happneing before.  I suspect that I updated to
> NetBSD 11 in May or June; I can't remember my last zfs lockup.
>
> I do have 12 GB in pools, but I suspect that's a lot of ZFS ARC that
> does get freed under memory pressure.
>
> Another 11 system that is my dom0 for pkg builders, also ZFS, after 24h
> of uptime (it's stable; I've just done a build/update/reboot pass.)
>
>   Memory: 2199M Act, 1097M Inact, 19M Wired, 110M Exec, 1948M File, 413M Free
>   Swap: 40G Total, 40G Free / Pools: 5818M Used / Network: 74K In, 150K Out
>
> A NetBSD 10 system that's been up 30 days:
>
>   Memory: 2476M Act, 1355M Inact, 17M Wired, 192M Exec, 1662M File, 311M Free
>   Swap: / Pools: 639M Used / Network: 40G In, 39G Out
>
> which is starting to be a pattern.
>
> So:
>
>   Is anyone else seeing excessive/growing wired memory?
>
>   Other than vmstat -s to see the wired page count, anything I should be
>   watching?
>