Sun Enterprise 6000, serial-number 725m9420
6th November 2019
Contents:
Views of The Whole System
The System in Action
A 90's Sun/Solaris system should always be demonstrated running the HotJava web-browser, and showing it's LEDs.
Hardware Status and Notes
- In the pictures, the external white component-labels do not
represent the usual Sun-style topological arrangement of the boards
("which board is in slot 3?"), but rather are intended merely to to keep track
of which exact module is which, for the detailed FRU list, below.
- As can be seen, this is a "standalone" E6000, without the 56-inch Datacenter Cabinet.
It has been fitted with heavy-duty castors, to enable easier relocation/movement.
Due to its' standalone nature, on power-up it does complain that it cannot detect
the Datacenter Cabinet Fans (of course). In the current configuration, and freestanding in the middle of a room,
neither the system nor
the CPUs get excessively hot (the CPUs, under load, run at 48C; the OBP/Solaris
CPU-monitor cutoff temperature is 85C).
- The non-Sun external CD-R drive and the high-half-termination SCSI data-cable have been unearthed
and re-tested with the system (used to install Solaris 2.6).
- Update, 5th November: fully reseating Power-Cooling Module #4 has resolved
the problem where POST previously detected the module as 'failing'.
All PCMs now consistently show good on POST and under load, even when running level-40 diagnostics.
- CPU/Mem board #6 is faulty. In particular, POST diagnoses
the first SIMM in both banks on the board with (correctable) ECC errors, and
the two CPUs on the board fail cross-call tests shortly afterwards. In fact, whenever this board is plugged in (into *any* Gigaplane slot),
one or more of the PCMs also freaks out and faults. Whenever this board is left disconnected, everything is fine, every time. Thus I suspect one of the DC-DC convertors on the board has failed badly (perhaps a short-circuit or open-circuit failure?). The pair of CPUs, and most or all of the SIMMs,
on that board may or may not be fine.
For now, I have left that board out of the system.
- The Gigaplane/centerplane, and its' slots, are good; proven by
swapping the above faulty board into different slots - the ECC/cross-call
faults *always* move with that board.
- The system has been fitted with brand new NVRAMs - one on the
clock-board and one on each of the I/O boards. The new NVRAMs have been
programmed with the original 1997 MAC address and hostid.
- With the faulty CPU/Mem board removed (ie: 16 CPUs running, 128 SIMMs and
7 Power/Cooling modules installed), the system can be powered from a
UK 240V domestic mains circuit via a standard 13A-fused plug. With all 18 CPUs
present, 144 SIMMs and all 7 PCMs, it was generally OK,
but did trip the protective mains supply
RCBO on one occasion, so perhaps the AC load fully-assembled in this configuration is right on the
limit of what a 15A domestic mains circuit can supply; on the other hand, the
AC-input-adaptor module of the E6000 is labelled "high leakage-current",
and thus the mains RCBO might have tripped due to leakage-current,
rather than current-overload; or even because the one faulty CPU/Mem board was
indirectly freaking the AC input (seems likely).
- Note that the Sun installation procedures and hardware manual for the
E6000 state that the inner chassis must be separately earthed, and although
that would of course be safer due to the AC leakage-current present, would
cause problems with a mains supply protected by an RCB or RCBO (which are
designed to break the circuit if the leakage-current to goes over
a small threshold).
- Solaris 2.6 and the latest recommended patch cluster is installed
on the 2.1GB disk on the disk-board. The 18GB other disk on the disk-board was
showing read-errors on a few sectors, and has been reconditioned using the
Solaris 'format' program, and retested with multiple passes of
format->analyze and dd; it now can reliably read and write every sector. However,
both disk drives are very old - their long-term health can only be guessed
at.
Link: Enterprise 6000 Hardware/System Manuals.
Selftest Verbose Diagnostics
The following reports were collected on 6th November 2019 (with the faulty CPU/Mem board #6 removed).
- POST/OBdiag, with
diag-level=max
.
- Verbose output of Solaris'
prtdiag
command, after a normal boot.
Full FRU/Module List and Pictures, by Group
16-Slot Logic Enclosure
Gigaplane CPU/Mem boards and their installed modules
Other Gigaplane boards and their installed modules
Gigaplane AC-DC Power Modules
External Keyswitch
Other