Re: OpenZFS: kldload zfs.ko freezes on i386 4GB memory

In message <20201030224734.GH2033_at_zxy.spb.ru>, Slawa Olhovchenkov writes:
> On Fri, Oct 30, 2020 at 03:34:10PM -0700, Cy Schubert wrote:
>
> > In message <20201030220809.GG2033_at_zxy.spb.ru>, Slawa Olhovchenkov writes:
> > > On Fri, Oct 30, 2020 at 01:53:10PM -0700, Cy Schubert wrote:
> > >
> > > > In message <20201030204622.GF2033_at_zxy.spb.ru>, Slawa Olhovchenkov write
> s:
> > > > > On Thu, Oct 29, 2020 at 08:13:00PM -0700, Cy Schubert wrote:
> > > > >
> > > > > > In message <E1kWvLj-0000GY-Ic.qroxana-mail-ru_at_smtp29.i.mail.ru>, qr
> oxan
> > > a 
> > > > > > writes
> > > > > > :
> > > > > > > 
> > > > > > > Hi,
> > > > > > >
> > > > > > > I have an old i386 machine running r364479. After upgrading to
> > > > > > > r367045, running kldload zfs.ko freezes the whole system.
> > > > > > >
> > > > > > > I also tried to replace the 4GB memory with another 2GB one
> > > > > > > and kldload zfs.ko works without freezing the machine.
> > > > > > 
> > > > > > ZFS ARC stresses memory. I've found a number of bad RAM chips over 
> the 
> > > > > > years using ZFS.
> > > > > > 
> > > > > > The OpenZFS upgrade significantly changed how it manages ARC. It's 
> like
> > > ly 
> > > > > > that prior to the OpenZFS upgrade your memory wasn't stressed to th
> e po
> > > int 
> > > > > > of failure. You can try to mask the problem by reducing your RAM cl
> ock 
> > > rate
> > > > >  
> > > > > > or or increase one of the other latency settings in your BIOS. Howe
> ver,
> > >  
> > > > > > again, this only masks an already weak RAM chip.
> > > > >
> > > > > Sounds like performance drop and regression
> > > > 
> > > > How so. Please explain.
> > >
> > > More stresses memory usually refers to performance penalty.
> > > Usually way for better performance is reduce memory access.
> > 
> > The reason filesystems (UFS, ZFS, EXT4, etc.) cache is to avoid disk 
> > accesses. Nanoseconds vs milliseconds.
>
> I mean compared ZoL ZFS ARC vs old (BSD/Opensolaris/Illumos) ZFS ARC.
> Any reaason to rise ARC hit rate in ZoL case?

That's what hit rate is. It's a memory access instead of a disk access. 
That's what you want.


-- 
Cheers,
Cy Schubert <Cy.Schubert_at_cschubert.com>
FreeBSD UNIX:  <cy_at_FreeBSD.org>   Web:  https://FreeBSD.org
NTP:           <cy_at_nwtime.org>    Web:  https://nwtime.org

	The need of the many outweighs the greed of the few.