Web lists-archives.com

Re: The Internet locks up Buster




On Thu, Jun 07, 2018 at 02:13:17PM -0400, Borden Rhodes wrote:
> > I.e. 12309 bug is back. It's obscure and presumably fixed (at least four
> > times fixed) bug that happens with relatively slow filesystem (be it
> > SSD/HDD/NFS or whatever) and a large amount of free RAM. I first
> > encountered the thing back in 2.6.18 days, where it was presumably
> > implemented (as in - nobody complained before ;).
> 
> Thank you, Reco and Abdullah, for providing some very helpful
> information. I'll retest with the kernel parameters. I went over to
> https://bugzilla.kernel.org/show_bug.cgi?id=12309 and it seems they've
> closed the bug and/or given up on this. Is there any value in
> continuing to whine about this problem? I mean, it's not like
> large-capacity RAM is going away.
> 

I feel like we are missing a trick here. Even with a relatively slow I/O 
device (I was faintly amused to see SSD in the list of relatively slow 
devices, if SSD is slow what is fast?) it should eventually catch up 
UNLESS something is generating an insane amount of I/O over a sustained 
period. Just browsing the web shouldn't do that unless RAM is very 
tight, and the O/P indicated they have lots of RAM.

I run my machine here with 24GB RAM and part of my filesystem is on an 
external USB hard drive cage. From reading this thread you'd think that 
when I run data analyses reading and writing that external drive cage, 
it would be a recipe for this bug, but it isn't. And that is because 
those processes do a lot of work and make the CPU work hard, but they 
don't do insane amounts of I/O. (lots, but not insane amounts)

So, I think the O/P should look into what is causing all the I/O in the 
first place, and why that I/O is sustained even when most of the 
processes on the system are blocked. Something isn't right there. The 
usual suspect would be swapping but again the O/P said they have 
"large-capacity RAM" and were just browsing the web with or without 
LibreOffice open -- this shouldn't trigger swapping.

Mark