NIS+ slowdown when servers are not reachable

[Peter Bunclark]

On Wed, 26 Sep 2001, Robert Greimel wrote:

> Hi,
>
> I have nis+ installed on Redhat 7.1 . Things are (mostly) working,
> but when one of the nis+ (replica) servers is down then we are
> experiencing considerable slowdown on the Linux boxes.
>
> For example, niscat passwd.org_dir returns immediately when all servers are
> up. When one server is down it takes 4 seconds to respond. When two
> servers are down it takes 8 seconds to respond, etc...
>
> This looks like an implementation problem.
I'd agree with that;  I get the impression Linux NIS+ almost `binds' like
yp, and doesn't seem to do the load-sharing which is part of the
advantage of using NIS+ over NIS (ie frequently choosing a different
replica).   It also can't find a sub-domain served from a different
server to the root server, which makes me suspect it doesn't really do
NIS_FIND_DIR properly...
 >
> Also, we have preferred NIS+ servers set for the Linux computers but it looks
> to me that this information is not being used ? Is this correct ?
> When will this be implemented ?
>
I would guess the concept of preferred servers came into Solaris round
about the time Thorsten stopped doing serious new development - although
he has certainly done some work more recently (bug fixes?).
	Please don't see this as a critism, it's remarkable that NIS+
works on Linux, but oh, it could do with some active development!

Pete.
> Greetings
>
> Robert Greimel
>
>