SPARC T3-1
 
SPARC SuperCluster T4-4 Full Rack
 
Oracle SuperCluster T5-8 Full Rack
 
SPARC S7-2
 
SPARC T3-4
 
Sun SPARC Enterprise T5220 Server
 
SPARC T7-1
 
Sun SPARC Enterprise T5240 Server
 
SPARC M7-16
 
SPARC T4-2
 
Netra SPARC S7-2
 
Oracle SuperCluster T5-8 Half Rack
 
SPARC T5-8
 
SPARC T8-1
 
SPARC T8-2
 
SPARC T7-4
 
SPARC T7-2
 
SPARC M8-8
 
SPARC T3-2
 
SPARC T5-4
 
SPARC T5-2
 
SPARC T8-4
 
SPARC T4-1
 
SPARC M7-8
 
SPARC SuperCluster T4-4 Half Rack
 
SPARC S7-2L
 
Oracle Exalytics T5-8
 
Oracle SuperCluster T5-8 Hardware
 
Sun SPARC Enterprise T5440 Server
 
SPARC T4-4
 
SPARC SuperCluster T4-4
 

PLA-Support>Sun Systems>SPARC>CMT>SN-SPARC: T5
 

Applies to:

Purpose

 This documents the source of power that Oracle recommends for it's T5xx0, T3-x, T4-x, T5-x, T7-x, & S7-x, M7-x, T8-x, & M8-x servers.

Scope

Details

   The power specifications for Oracle T5xx0 & newer servers are stated in the Released Installation Guide for each server, such as the SPARC T4-1's E22988-08.  The "Electrical and Power Specifications" section will state the input voltage, current, & power specifications.  DC powered systems typically accept voltages as low as 48V DC & typically recommend a positive ground (eg -48V DC).  AC powered systems could either allow 100V AC to 120V AC or 200V AC to 240V AC.  For example, the T3-2 minimally requires 200V AC.  Line voltage drop must be taken into account for power distribution with in racks for the system furthest from the source due to distribution losses.

Power Outage/Glitch Times:
   Oracle systems are designed to provide continuous operation on a power system that prevents simultaneous power dropouts of more than 10 milliseconds (typically 12 ms) on all power input circuits.   The following results are typical for simultaneous input voltage dropouts:

• Dropouts of 10 milliseconds or less do not impact system operation.  Errors may not be logged for some systems as discussed below.
• Dropouts of 10 to 40 milliseconds possibly result in a host shutdown and a fault recorded in the logs.
• Dropouts of more than 40 milliseconds could possibly result in loss of standby power so the Service processor (ILOM) could reboot.  A reboot will usually not provide enough time for the error to be logged.

The higher the PSU's load, the quicker it will drop out when power is lost.

Power Redundancy:
   To over come power interruptions, Oracle recommends 2 independent power sources.

   In an N+1 configuration such as the SPARC T4-1, if one PSU fails or needs service, then the remaining PSU is capable of maintaining normal system operation.  The same applies to loss of input power to the PSUs so power feeds to them should be derived from at least two different sources.  If a circuit breaker trips or there is an interruption in one power source & the other source is unaffected then the server will operate properly due to this redundancy.  Released Oracle documentation typically states the need for redundant power sources in the Server's Getting Started Guide, such as the T4-1's: E22992-02.  The server will continue to operate if one of two power sources fails if the redundant PSUs are properly split between them.  The redundant PSUs in Oracle systems typically share the load with minor variations between their output currents.

Uninterruptible Power Supplies (UPS):
   Lightning strikes & temporary power interruption can cause the server to crash unless one or more UPSs are in use, so they are highly recommended for critical systems.  Each UPS contains a battery or other energy storage device and the run time in backup mode varies from milliseconds to hours depending on the energy stored and the UPS load.  It is highly recommended that two UPSs be used (one for each redundant power circuit).  A UPS could fail or need service during the system's life so a second still provides power to the redundant power circuit. 

   UPSs have two commonly used operating modes.  One method is Standby USP mode where the UPS can switch to battery output mode if the input source falls out of threshold.  This mode could have problems when the UPS switches back to the AC source shortly after AC power is restored.  Some brands of UPSs may not properly handle several power interruptions over a short time period which could results in a power interruption to the server.  A more reliable method is to use Double Conversion On-Line UPS where the UPS output is always derived from a power inverter which is typically supplied by the AC input unless an AC problem occurs. The other modes are not commonly used, so usage should only be done at the UPS vendor's recommendation.

   A second UPS on the redundant power source helps to avoid outages at sites with problematic power sources since it's unlikely that the timing of the UPS output interruptions on both sources will overlap for 10ms.  Oracle recommends a pure sine wave UPS using the double conversion mode which provides the greatest reliability.

Please note that Schneider has a great document on the advantages of different operating modes.  You can point customers to this document by:

1. Access the Schneider URL: http://www.apc.com/site/support/index.cfm/faq/ .  
2. Enter the following into the search field: "Different Types of UPS Systems". 
   
3. The first document listed is: FA157448 - Different Types of UPS Systems.

SPARC T3-x through T8-x / M7-x through M8-x Server PSU generations & input voltage sensors:
  PSU's provides standby (3.3V or 12V on newer) DC to power the ILOM & other continuously powered components when input power is present.  It also supplies 12V DC to the host when the user powers the host on.  The 12V DC power to the host may be lost if all PSUs experience a simultaneous input power loss for more than 12 milliseconds.  The system ILOM and other components that operate in the standby mode can normally maintain operation for a simultaneous input power loss of 40 milliseconds or more.

A249:  Used in the T5x20, T5140, T5240, T3-1 & T4-1 systems.  The only PSU input sensor is VINOK which is periodically sampled at roughly a 5 second rate & is not latched.  VINOK indicates if the input voltage falls outside the upper or lower critical threshold even if the PSU remains operational.  Unfortunately, a sensor event is not typically detected if a PSU's power source has a temporary brown-out unless caught by a 5 second sample.    If a fault is detected, it is latched until cleared by the ILOM or by the PSU being power cycled. 

   If all PSUs lose power at once, then there is no indication that the system rebooted due to power loss & a POK fault is typically generated for the event.  When A249 PSUs power the system & a host reboot occurs, one should first determine if the also ILOM went offline & rebooted once power was restored.  An ILOM outage will always result in an ILOM reboot once power is restored.  On the other hand, a similar scenario can occur if the ILOM becomes inoperable due to a firmware problem which also affects the host & a power cycle is used to restore operation.  One should check for VINOK events on multiple PSUs with in the last 3 months which increases the likelihood that a reboot could be due to power loss.

A238/A246: Used in the T5440.  Same comments as the A249.

A239:  Used in the T3-2, T3-4, T4-2, T4-4, & T5-2.  These PSUs have 2 latched sensors:

• V_IN_ERR  is latched when the input voltage falls outside either of the critical thresholds.  Unfortunately, this register is cleared when the PSU power is recycled.
• V_IN_WARN (on T5 servers) is the result of the Power Supply under voltage register being set when the input voltage falls outside either of the warning thresholds.  The warning from this sensor was changed to an informational message in firmware version 9.5.

If all PSUs lose power at once, then there is no indication that the system rebooted due to power loss.  When A239 PSUs power the system & a host reboot occurs, one should first determine if the also ILOM went offline & rebooted once power was restored.  An ILOM outage will always result in an ILOM reboot once power is restored.  On the other hand, a similar scenario can occur if the ILOM becomes inoperable due to a firmware problem which also affects the host & a power cycle is used to restore operation.  One should check for V_IN_ERR events on multiple PSUs with in the last 3 months which increases the likelihood that a reboot could be due to power loss.

  T4-2 & T4-4 systems with firmware 8.2.2.b or newer may not show most V_IN_ERR events due to event filtering added to that firmware.

Please see T4-2 issue 10: I2C bus filtering due to FCO: 1570318.1.

A258: Used in the T7-1.  Firmware supplies PSU register info on some PSU faults via ereports.  A power cycle clears these registers, though.

A261: Used in the T5-4, T5-8, T7-4, M7-8, & M7-16.  Same comments as the A239 unless firmware 9.5.1.b is installed which allows PSU register data to be displayed in some PSU fault related ereports.  This data remains intact after power cycles & is cleared via system clear.  The M7 PSU's firmware can be obtained: here.

A263: Used in the T7-2 & T8-2.  Firmware supplies PSU register info on some PSU faults via ereports.  This data remains intact after power cycles & is cleared via system clear.

A265 (Bell) & A265D (Delta): Used in the T8-4 & M8-8.  Similar to A261 but different size & input faults retained over power cycles.

A266: Used in the T8-1.  Similar to A256 & A258 but double the output power, input faults retained over power cycles, & additional functionality added.

FMA Faults related to input power:
SPT-8000-5X (fault.chassis.env.power.loss) - Power Supply AC Input Voltage Failure
This fault is not generated often for A249 based systems.  The customer/admin should check AC inlets, power sources and determine the reason for the loss of AC power and correct the problem with unreliable power.

SPT-8000-MJ (fault.chassis.power.fail) - Power Supply general failure
Fault can occur for a variety of reasons including AC power, AC Power cord unplugged, etc. Admin should verify air inlets are not obstructed or another condition is not causing this event 

FMA Alerts related to input power:
SPT-8002-C4 (alert.ilom.chassis.power.loss) - AC input failure
This Event can occur on SPARC S7 and SPARC T7 systems.

This solution has no attachment