On Thursday 24 September 2015 05:32:53 Richard Glock wrote:
On Thu, 24 Sep 2015 05:18:52 Gene Heskett wrote:
On Thursday 24 September 2015 03:03:10 Dr.
Nikolaus Klepp wrote:
> >
Subject says it all. I need to find the experts.
> >
> > Cheers, Gene Heskett
>
> I use my local Linux User Group, full service.
My local linux user group. Chuckle. I am 1 of a group of 3. Not
too many linux users in these here parts. I am quite likely 100
miles from the nearest "user group" that numbers 10 or more.
> I use nfs on my local network, it just works so I am far from
> an expert. I export my "/home/<user>" dir and manually mount ,
> cli, on the clients.
>
> Debian stable.
Debian Wheezy. With TDE.
Cheers, Gene Heskett
Hi Gene!
I dropped NFS on linux ages ago, due to simillar issues as you
describe. Now I use SSHFS and haven't had any issues since then.
So, what about using SSHFS instead of NFS?
Nik
Never heard of it till now. So I installed it, along with sshmenu
which pulled in a dozen other rubyish packages.
Silly Q though, does mc understand sshfs? Or do I need to find a
new 2 pane file manager that does understand it?
One thing's for sure, NFS, even V4 is old enough to have bit rot.
Thanks Nik. Off to read some man pages.
Cheers, Gene Heskett
On the MC command line, do cd fish://user_name@machine_name
Resource temporarily unavailable, fish might not be installed?
But a "cd /sshnet/shop", after using Niks's example sshfs command, a
mount point I created, then chown'd to me, works just fine. Since I
also have an ssh -Y session into each of those machines, if I need to
muck around out of my home dir, sudo is always available.
To summerize, I added these lines to /etc/fstab:
shop.coyote.den:/ /sshnet/shop fuse.sshfs defaults,idmap=user 0 0
lathe.coyote.den:/ /sshnet/lathe fuse.sshfs defaults,idmap=user 0 0
GO704.coyote.den:/ /sshnet/GO704 fuse.sshfs defaults,idmap=user 0 0
Which I suspect can be nuked. but there it is. The mount points were
created and chown'd to me.
Now I will see if the fstab entries are surplus by doing the same thing
to each of the other 3 machines currently alive on this local network.
Then I can hopefully reach across the net from any machine to any
machine, which was my target in the first place.
According to my results of doing the mkdir yadda, followed by the sshfs
login, it works just fine on GO704. I can look at /home/gene on this
machine from the ssh -Y session into that machine. Two more machines
to go... But first, clean up the mess in my fstab.
Oh, and sshmenu is broken, needs a ruby dependency the deb didn't list. I
don't have a heck of a lot of ruby stuffs in use here. I'll nuke it.
Nik's example sshfs command line was then executed once for each of the
mount points.
Humm, on GO704 it all works, and here it all works BUT the sshfs session
converts the individual subdir used from gene:gene to root:root.
If thats permanent it will be a problem.
So I go over to the ssh -Y session into the lathe and do the mkdir
tapdance again. But while it will connect to both this machine and
GO704, it will not connect to "shop", "connection reset by peer", so
once again that shop machine is being a spoiled brat.
Doing that same tapdance on machine "shop" works as expected.
Now why can't lathe access shop?
gene@lathe:~$ sshfs gene@shop:/ /sshnet/shop
read: Connection reset by peer
However, gcode written for the lathe (only a 2 axis machine) is not
usable on shop or GO704 which are 4 axis milling machines, so that is
not a showstopper loss. Besides, I can go to the other machine and do
the file copies if I need it bad enough.
What does bother me though is that if the ownership
of /sshnet/machinename being changed to root is permanent, that will
mean I have to do the "sudo chown -R /sshnet" dance on 4 machines when
they have been rebooted. That is the only way I know to get around the
target machines asking me for a non-existent root pw.
NFS these days is a hairball of epic proportions. Try
getting the NFS
daemons to bind to a specific address per the man pages...
And has been so for at least 5 years, the level of neglect seems rampant.
The manpages haven't been touched in 9 years.
But till now I have used it if I could because there wasn't a usable
alternative. I long ago got tired of the constant perms fixing that CIFS
needs, too many roots in the M$ world.
RG
Many many thanks to both Nik and Richard for supplying the clues and
examples that made it work . And to Timothy for pointing out that it
might be a problem with rpcbind. But yesterdays huge update included
rpcbind, which all machines have now been updated, and that did not fix
nfs.
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>