Gilles Schipper wrote:
> 1. You MUST ensure the "number of non-nailed terminals", in the "Host
> Configuration Screen" of NMMGR is greater - preferably equal to the maximum
> number of telnet and/or non-nailed VTMGR sessions you expect to have
> connected to this HP3000.          ^^^^^

This could vary by release, but as far as I know, "VTMGR/[host]telnet"
sessions are allocated as VTERMs dynamically at connection time and are
in no way related to "non-nailed terminals" in NMMGR.  If you examine
your configuration in SYSGEN (after a RDCC cross-check) you will find
nailed devices by doing "io> ld xxx/yyy" across your DTC ports, or if
you don't specify device numbers, lists all "sysgen-aware" ldevs.

Immediately after a boot, if you do a "showdev" you will find a quantity
of ldev numbers (unused ones from sysgen) allocated beginning at your
first unused ldev number.  These are your "non-nailed terminals" and
they are pooled for use by DTC connections (from non-nailed ports or
Telnet Access Cards (TACs)).  The NMMGR "non-nailed terminals" value
determines an upper bound of these connections.

If you go a step further and note the first unused ldev from the display
above, this will mark the starting point for the VTERM pool.  VTERMs
are dynamically created at connect time and released after logoff; in
other words you can :showdev the device when in use, but it disappears
as soon as you disconnect.  On 5.5, VTERMs appear to be used (in a most
peculiar way) for /dev/tty device links.  Higher levels of LISTFILE of
such a device link will show a VTERM ldev (which remains constant,
despite the dynamic nature of VTERM allocation, but I'm off topic...)

As a sanity check here (I remember this because we used ldev ranges for
security checks to determine a session's origin prior to the CI vars
available in 5.5) the first incoming non-nailed DTC/TAC session gets a
low ldev number (usually the first unused one on the system).  The first
incoming NS/VT or host telnet session gets a higher number.

There is no obvious upper bound to host telnet VTERMs.  You can limit
NS/VT (VTMGR) sessions with NSCONTROL by assigning a max number of
servers, but there's no such animal for host telnet that I'm aware of.
I posted some workarounds in a previous post (check the archive, I
don't want to repost the lengthy reply here).

Jeff Kell <[log in to unmask]>