Will turning off the
HP3000 system console lock up the system?
BT
Today, the system doesn't care if the console is switched on or not. However, if the
console is left hung via a control-s, console message
buffers will eventually fill up and any process that tries to send a console message will
be suspended until the message can be delivered, at
least for a minute or so.
MP
There are 63 what I call genmsg buffers. Every console message requires a genmsg buffer
and a TELL or WARN requires
a buffer per destination terminal. When the system detects that all of the buffers
are in use, it starts a 30 sec. timer. If no buffers free up after
that time, the system issues a hard-preemptive I/O to the console in an attempt to free it
up. Then, the system waits another 30 sec. If the
console is still hung, then we abort every I/O that's using a genmsg buffer.
So, if the console can never complete I/Os, for whatever reason, then there is a 63
message/one minute timeout/abort 63 I/Os cycle that continually repeats itself. A
system with a lot of console/TELL/WARN I/O traffic will seriously bog down as the result
of a hung console.
BT
After that, the system gives up and starts tossing all its undelivered messages into the
system log file. So if you _have_ managed to hang up your
console (turning it off won't do it by itself), you'll need to look in your log files. I'm
not sure what happens if you don't have console logging
turned on -- the system won't crash, but it's not clear what, if anything, will happen to
undeliverable messages.
MP
Actually, the system doesn't do anything special with the aborted I/Os. If console logging
is enabled, then the aborted console messages will be found in the console log files.
If console logging is NOT enabled, then these messages are lost. If the
aborted message was targeted at a non-console terminal, then this message is always lost.
S/68 - S/70 --- Increased CPU (not disc) Cache for a rather significant increase in
CPU speed.
There were a couple of other things that differentiated the S/64A and S/68A from the
S/64B, S/68B, and S/70. Most notable were a different
SSDP (System Status and Display Panel) and a totally different power supply and
distribution group.
JK
(As I remember) the offerings at the time were 40, 44, and 64. The 40 and 44 were
essentially the same horsepower and had interchangeable boards, but the 40 was a
square-ish box with no terminal connections (other than console); the ADCCs ran out to
little metal boxes or your ATP came to a mini-cabinet. The 44 was the "desk"
looking thing, the 64 was the "chest freezer".
MPE V started the cache business and had V/P initially (cache only), V/E (new processor
and firmware for CST tricks), and later V/R (for the obsolete Series III and below).
V/P was purely software, but you had to "upgrade" from 40-42, 44-48, or
64-68 to get it; consequently if you got the upgrade early (before the firmware was
released) and deleted the memory board, all you got was a new FOS/SUBSYS tape and a new
nameplate (for a considerable chunk of money). I know, I got two of them :-)
Of course you later got the firmware (and V/E).
The next processor speed-up changed 42-52, 48-58 (for sure) and I think also 68-70.
I've had a 40/42/44/48/52/58, dunno about the rest :-)
CB
Yes, 68-70 added CPU/microcode caching. We went through a few of those upgrades
ourselves.
|