[rescue] Sol10 U4 info from Sun

[N. Miller]

On Aug 15, 2007, at 1:03 PM, Meelis Roos wrote:

 > On Aug 15, 2007, at 11:54 AM, Bill Bradford wrote:
>> NIS+ was just created for a specific LARGE customer that needed it.
>
Do tell, Bill.

> Well, we (or rather they) tried to use it at Tartu University,  
> Estonia.
> During some first Solaris releases it was weak and unstable for us  
> (plus
> the learning curve). Like adding a backup server when the primary went
> down - should have helped, but it actually took down the secondary
> server alogn with the primary. Sometime around 2.4 or 2.6 it became
> stable and our admin got good-enough grasp of it - but the some other
> departments were added (around 9000 accounds and around 10 subdomains
> total) and it seemed we grew out of it again. Like a race, we adding
> users and Sun fixing NIS+. We seemed to be a too big customer, and  
> that
> was strange - a simple university.

You haven't seen the serious mess it can make.  We had a Sun SE *on- 
site* for 30+ days at no cost to $network_juggernaut around '97 time  
frame, after no engineer (~3K or so) could log in to their  
workstations, engineering servers, etc. for 3+ days thanks to NIS+  
melting down.

The secondary servers were across slow links and the NIS+ masters  
just could not cope with the time it took for updates (all engineers  
having to be able to log into all servers in all time zones, don'tcha  
know?).

It was the epic wreck of poorly designed (and untested, I might add)  
architecture destroying software that couldn't scale.  I was just  
glad I was a near n00b SA at that point and not involved in the hell  
weeks that followed other than assisting in the work around's  
deployment on the servers I supported in India.

Ultimately, the master dbs were replicated and pushed around by  
external processes on a daily basis outside of the NIS+ system as the  
workaround.

The Sun SE later confided in us that even Sun didn't use NIS+ in  
production.

Oh, and it happened the 36 hours before fiscal year division results  
were announced.  The engineering support director's 15 minutes in  
front of the engineering VPs, directors, managers, and employees were  
quite satisfying in a perverse way.  He was an idiot, and probably  
the worst of the long string of bad directors we had in that department.

=Nadine=