SGI Tezro Troubleshooting Guide

Tezro front with the drive shroud off: the L1 display reads "001c01 Prom Running" above the lit bezel LED (Telecineguy, CC BY-SA 3.0)

Fault-finding on the SGI Tezro, from the troubleshooting and diagnostics chapters of SGI's tower and rackmount hardware guides and the L1 controller guide, with repair experience from the IRIXNet forum. The Tezro is supervised by its L1 system controller, which starts as soon as mains power is applied, runs the fans, watches voltages and temperatures, and powers the system on. Most diagnosis starts at the L1 display on the front or the L1 console port at the rear.

Bezel LED

Tezro bezel LED signals (tower Hardware User's Guide, Table 4-1)

LED	Meaning
Blinking white	Power button pressed (on or off)
Solid white	PROM booted successfully or operating system running
Solid yellow	The L1 has detected a problem; read the L1 display
Blinking red	General system failure; read the L1 display
Solid red	System node board failure: the PROM could not be read at power-on

Environmental shutdowns

The workstation powers itself off if any fan runs below 80% of its rated speed, any temperature sensor reaches 70 °C (158 °F), or any voltage moves 20% from nominal. SGI's rackmount guide explains the L1 messages behind these: a voltage 10% from nominal gives a warning and 20% a fault, a fault message starts a 30-second power-off sequence, and temperature messages escalate from advisory (fans at 80%) to critical (6 °C over target, fans at 100%) to fault (10 °C over target).

L1 messages (rackmount Hardware User's Guide, Table 4-2)

Message	Meaning and action
ATTN: x.xV high (or low) fault limit reached @ x.xxV	30-second power-off sequence
ATTN: x.xV high (or low) warning limit reached @ x.xxV	Voltage outside nominal
ATTN: FAN # x warning limit reached @ xx RPM	A fan has increased its speed; check the room temperature and the clearances
ATTN: FAN # x fault limit reached @ xx RPM	A fan is below its minimum speed; check whether it has failed
ATTN: TEMP # advisory / critical / fault temperature reached @ xxC xxF	Lower the room temperature, clear the front and rear, check for blocked airflow
Auto power down in xx seconds	The L1 has registered a fault and is shutting down

Use the L1 env command to see each fan's speed and each voltage and temperature against its limits, and log for the history of messages. IRIXNet members report that the L1 also prints "INFO: Cannot enable VRM: 9", 10 and 11 at power-up on a Tezro and that these are normal: the L1 firmware is shared with systems that have those regulators.

L1 does not start

With mains connected, the L1 should start, show its display and run the fans. If the display backlight comes on but nothing is shown and the fans do not start:

1. Connect a terminal to the L1 console port at the rear with a null-modem cable and see whether the L1 answers. IRIXNet members suggest 38,400 baud.
2. Unplug the supply's 24-pin and CPU power connectors from the system and test the supply with a hand-held ATX tester. The L1 needs good 3.3 V and 5 V rails. A Tezro supply reads low on +12 V1 (11.1 to 11.4 V) and shows +12 V2 as low on such a tester, which is normal; a low 5 V is not, and was the fault in this case.
3. Check that the Dallas clock module used by the L1, a black module between the PCI slots and the fan wall, is seated. In this case the L1 only started once both the supply and the Dallas module had been replaced with known good parts.

Also check that the power button is not sticking.

Serial number and NVRAM errors

• "SYSTEM SERIAL NUMBER INVALID FORMAT" and an immediate power-off after the Dallas module is taken out or replaced: the module has lost the serial number. Set it again from the L1 using the serial number on the back of the system, then power up. The L1 serial command shows the stored values.
• "NVRAM checksum is incorrect: reinitializing" after every power cut: the battery of the SNAPHAT on the IO9 board, which keeps the PROM environment and the IRIX time of day, is flat. In one IRIXNet case a new M4T32-BR12SH6 SNAPHAT stopped the error.
• If both clock batteries are replaced at once, the system cannot copy the serial number from one to the other and has to be told it by hand.
• With a new Dallas module the L1 powers the system up by itself about a minute after mains is connected. IRIXNet members say this automatic power-on is the default on the Tezro and Fuel.

Power-on diagnostics and PROM

The PROM tests the processors, the Bedrock ASIC, the PROM itself, the DIMMs, secondary cache, the PIC ASICs, the PCI slots, the serial ports, the SCSI controller and the VPro graphics in that order. A failing part is disabled and testing continues; the system then may or may not boot. In one IRIXNet case, a Tezro stuck with "General Exception" messages in the PROM turned out to have two failing DIMMs and a fault that appeared only with the keyboard plugged in. With the two DIMMs removed and no keyboard connected it reached the PROM menu on the serial console; the keyboard fault was still unresolved when the thread ended. Members there suggest reducing the system to the node board, the IO9 board and one DIMM pair, without graphics or keyboard (the PROM then uses the serial console), before suspecting the interface board. After changing hardware, run enableall and update at the PROM prompt so that it stops reporting the old inventory.

Offline and online diagnostics

• Offline: choose Run Diagnostics from the System Maintenance menu, or type boot -f dksc(0,1,0)/stand/smdk/smdk --a at the command monitor. The launcher tests the CPU (single-processor systems only), secondary cache, DIMMs and the IO9 and daughtercard I/O, and stops at the first FAILED result. SGI gives about 6 minutes for a full run.
• Online: as root, run /usr/diags/bin/runalldiags with -basic, -normal (default) or -extensive. Run it only on an idle system, from the console for the normal and extensive modes; failures are saved in /tmp. Unplug and reconnect USB devices afterwards.

No picture

On a VPro board without the Dual Channel Display, the PROM screen is sent as analogue video and IRIX uses the digital signal, both on the DVI-I port. IRIXNet members report that a DVI-D cable, or an adapter to HDMI or VGA, loses one or the other; use a DVI-I cable to a monitor that accepts both. A long pause with a black screen during boot can also be the network configuration timing out.

Power supply faults

The tower supply provides 3.3 V and 5 V at 24 A each, three +12 V outputs and 5 V standby, with 500 W combined. IRIXNet members who rebuild these supplies describe them as ATX-like: the 24-pin connector follows the ATX layout, but the 12 V readings on a tester are low by design. A supply that cannot be repaired can be replaced with an ATX unit wired to the Tezro's 10-pin connector; see SGI Tezro General Maintenance. weblacky also reports a Tezro whose 1.8 V regulator output sags at start-up, which he attributes to damage to a regulator circuit on the main board, not to the supply.

Related pages

• SGI Tezro
• SGI Tezro General Maintenance
• SGI Tezro Capacitor Replacement Guide
• SGI Tezro Documentation
• Capacitor Failure Symptoms