Web lists-archives.com

Re: Why choose Debian on server




On Friday 04 January 2019 14:31:10 deloptes wrote:

> Gene Heskett wrote:
> > Can I trade the pickity nfs link in on an sshfs version? I use that
> > here as its quite bulletproof. The login without the checksum files
> > might be a hassle though. I'll certainly look that link over, thank
> > you.
>
> I don't know about sshfs - not heard that you can mount root over
> sshfs. if on NFS root it speeds up testing RPI a lot. After you get
> the magic combination working good, move the setup to the SD card if
> needed.

No, or at least I've not done it, if I have something that needs to go in 
a root user controlled area, I copy it to my /home/me partition on the 
target machine, then ssh in as me, and sudo to get the rights needed to 
put the file where it needs to go.  So thats not something readily 
amenable to scripting. That would be handier than bottled beer at times, 
but I can also understand the security problems that would expose if it 
blindly allowed the copy. Its been years since I had an NFS setup, seems 
it was always on the missing list whenever I needed it the worst. Within 
its perms limits, sshfs Just Works(TM).

> As for the kernel I build with cross-compiling toolchain and used qemu
> to the nfs share. The RPI2B was really slow.

So is the 3b, since the SSD I have attached so I have room enough to 
play, must also be accessed thru that internal and slow usb-2 port.
I build those kernels on the SSD, not on the u-sd using up all those 
write cycles. On the rock64, similar SSD plugged into the usb-3 port, 
which due to a silicon bug, has been resticted to usb-2 speeds, but its 
about an hour on the rock64 to build the same but arm64 flavored kernel 
on it. I can live with that if it actually made it usable.

Gawd I miss grub on these teeny power mizerly things.
>
> I look now at the notes and it feels I see them for the first time ...
> it was just 4y ago.

At my age, 84, its a lot quicker than that. Some vitamin b1 seems to help 
me pay attention though.

> I am very interested in the RT issues you discuss, but mostly because
> I know that linux is not RT and couldn't have time to investigate,
> however it is one problem for the industry/machinery.

And becoming a bigger problem by each kernel release.  Just the bigger 
stack image a context switch involves takes the 64 bit stuff into the 
very close to unusable state. Even the hit of enabling pae on the 32 bit 
stuff is a quite noticeable hit on the rt performance. And it doesn't 
make a lot of sense, with 2 gigs of ram in one of these machines, it 
rarely gets into swap. Put another 2 gigs in it, and before you can say 
uptime, its into swap by 3-5 hundred megs. Pure nonsense to this old 
fart. Sigh, but they won't let me anywhere near the rope that sounds the 
whistle.

> regards


Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>