[comp.sys.apollo] Problem with groups

rn@ap.co.umist.ac.uk (bob nutter) (03/19/91)

Hi!

Has anyone experienced the following, and do they have any solutions?:-

On certain nodes, an su is rebuffed with `You do not have permission to su
root'. The thing is, I *do*! Running groups(1) just prints a blank line, whilst
doing 'groups rn' gives the expected answer. Thus what I have to do to su root
is first to su to myself, and then su to root. This baffles me. Perhaps I've
missed something obvious. The only thing that happened before this started was
that I ran cops and tightened up things quite a bit ;->. It's obviously a
registry problem (aren't they all?) so any help would be appreciated. This is
for 10.1, don't have the diskspace to go to 10.3 yet...

bob 
-------------------------------------------------------------------------------
bob nutter, computer officer    | "If there's no class divide, then how come
UMIST dept of computation       | you're standing here waiting for a bus
po box 88 manchester m60 1qd uk | when somebody drives past in a car worth
tel:+44 61 200 3386             | the price of a house?"
email:b.nutter@umist.ac.uk      |     -Class War grafitti

etb@milton.u.washington.edu (Eric Bushnell) (03/21/91)

In article <1991Mar19.154313@ap.co.umist.ac.uk> rn@ap.co.umist.ac.uk (bob nutter) writes:
>
>On certain nodes, an su is rebuffed with `You do not have permission to su
>root'. The thing is, I *do*! Running groups(1) just prints a blank line, whilst
...
>for 10.1, don't have the diskspace to go to 10.3 yet...

Yes! But I get it at 10.3. groups(1) usually returns an *incomplete* list.
Earlier discussion leads me to believe that rgyd does *not* update
/etc/passwd[group] in a timely or accurate way. 

I've also had trouble with reusing unix id numbers. I assigned a number
from a deleted acct to a new user, and *one* machine insisted upon 
using the previous owner's name and rights.

Could we force rgyd to update the unix files if we delete /etc/passwd[group]?


Eric Bushnell
Univ of Washington Civil Engineering
etb@zeus.ce.washington.edu
etb@milton.u.washington.edu

wjw@eba.eb.ele.tue.nl (Willem Jan Withagen) (03/21/91)

In article <1991Mar19.154313@ap.co.umist.ac.uk> rn@ap.co.umist.ac.uk (bob nutter) writes:
=>Hi!
=>
=>Has anyone experienced the following, and do they have any solutions?:-
=>
=>On certain nodes, an su is rebuffed with `You do not have permission to su
=>root'. The thing is, I *do*! Running groups(1) just prints a blank line, 
=>whilst
=>doing 'groups rn' gives the expected answer. Thus what I have to do to 
=>su root
=>is first to su to myself, and then su to root. This baffles me. 
=>Perhaps I've
=>missed something obvious. The only thing that happened before this 
=>started was
=>that I ran cops and tightened up things quite a bit ;->. It's obviously a
=>registry problem (aren't they all?) so any help would be appreciated. 
=>This is
=>for 10.1, don't have the diskspace to go to 10.3 yet...
=>

Did you by any change put yourself into more than 8 (eight) groups.
I once did that, and it had the same type of effects. ( Groups(1) ... )

The problem could then lie in the original BSD code which allows a person
to be member of no more than 8 groups.
Eceeding this limit is asking for trouble.

Regards,
	Willem Jan Withagen.
Eindhoven University of Technology   DomainName:  wjw@eb.ele.tue.nl    
Digital Systems Group, Room EH 10.10 
P.O. 513                             Tel: +31-40-473401
5600 MB Eindhoven                    The Netherlands

troy@plod.cbme.unsw.oz.au (Troy Rollo) (03/22/91)

From article <18743@milton.u.washington.edu>, by etb@milton.u.washington.edu (Eric Bushnell):
etb> In article <1991Mar19.154313@ap.co.umist.ac.uk> rn@ap.co.umist.ac.uk (bob nutter) writes:
rn>
rn>On certain nodes, an su is rebuffed with `You do not have permission to su
rn>root'. The thing is, I *do*! Running groups(1) just prints a blank line, whilst
etb> ....
rn>for 10.1, don't have the diskspace to go to 10.3 yet...

etb> Yes! But I get it at 10.3. groups(1) usually returns an *incomplete* list.
etb> Earlier discussion leads me to believe that rgyd does *not* update
etb> /etc/passwd[group] in a timely or accurate way. 

Which in itself wouldn't be a problem, as the last data to be put there should
be still there. So unless you've just been added to wheel, you will already
appear in /etc/groups. Unfortunately the Apollo functions to get password file
information seem to always refer to the registry, and if they can't find it,
give up. Perhaps a better recourse would be to then look in /etc/groups, but
Apollo would probably call that a weakness.

This can also be (and in my experience usually is) caused by your node using
somebody else's glbd for a while. If this is happening, and your glb is on a
node on the same subnet as somebody else's glb, you can find out as follows:

	# /etc/ncs/lb_admin:
	lb_admin: l
		.
		.
		.
You may find that this shows registries from other sites on your subnet.
This problem only occurs on gateway machines, as far as I can tell, and
is fixed by invoking glbd with -li dds (/etc/rc) and by listing the sites
where you expect to find your glbd in /etc/ncs/glb_site.txt in the form
dds://hostname

___________________________________________________________
troy@mr_plod.cbme.unsw.oz.au