Closed Bug 1575042 Opened 7 years ago Closed 7 years ago

verify/set moonshot power redundancy

Categories

(Infrastructure & Operations :: RelOps: Hardware, task)

task
Not set
normal

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: dhouse, Unassigned)

Details

The moonshot chassis in mdc1 will lose half of their power next week on Wednesday for maintenance. It looks like they are not configured for power redundancy: " Power Mode Mode 0 - No Redundancy". We need to confirm the configuration and make sure they are set to be redundant. We can also test the night before to verify that it handles the 1/2 power loss.

The options for power-mode are (https://support.hpe.com/hpsc/doc/public/display?docId=c03707430#_OPENTOPIC_TOC_PROCESSING_d94e9301):
<option>0 - No Redundancy mode</option>
<option>1 - Max Performance with redundancy mode</option>
<option>2 - Full AC/DC redundancy mode</option>
<option>5 - Power feed redundancy mode</option>

Power capping modes
The HPE Moonshot 1500 Chassis supports different power capping modes. The setting impacts how thechassis reacts to changes in power supply functionality, such as loss of input power to a power supply orfailure or removal of a power supply from the system. Selecting the appropriate power capping mode dependson system requirements, how the chassis is provisioned, and how power feeds are supplied. The powercapping mode alone does not guarantee availability in the event of power loss. The chassis requires at leasttwo power supplies. The type and number of power supplies depends on the cartridges installed in thechassis.The following power capping modes are standard and are configurable in the power management controller:
• Mode 0: No RedundancyAll power-capping is disabled. This mode can be used to minimize any possible performance impact ofpower-capping logic.
• Mode 1: Max Performance with RedundancyThis is the default power capping mode. This mode allows the maximum number of nodes to run byengaging power-capping if the power draw from the chassis attempts to exceed the load supported by theactive power supplies. In this mode, the system is expected to survive (with degraded performance) anunexpected power loss to one or more of the power supplies.
• Mode 2: Full AC/DC Redundancy ModePower-capping is enforced such that system has N+1 power redundancy. In this mode, the system willthrottle the nodes allowing for one of the available power supplies to be held in reserve. If only one powersupply is available, the system will throttle the nodes as to allow the full use of that power supply’scapacity.
• Mode 5: Power Feed Redundancy ModeWhen used with an A+B power feed configuration, Power Feed Redundancy Mode throttles the system100%, bringing the nodes to a complete stop if a power feed loss is deduced. Full throttling continues untilthe power feed is brought back online. In this mode, the system is expected to survive an unexpected lossof an entire power feed to half of the power supplies.

Q, do you know with the power redundancy for the moonshots, do we need to change any settings for 1 of the 2 power sources going down for maintenance/replacement next week Wednesday (8/28)?

To be safe, it looks like we can set the power-throttling in case we reach the limit when we have only 1 power source:

(In reply to Dave House [:dhouse] from comment #1)

• Mode 1: Max Performance with RedundancyThis is the default power capping mode. This mode allows the maximum number of nodes to run byengaging power-capping if the power draw from the chassis attempts to exceed the load supported by theactive power supplies. In this mode, the system is expected to survive (with degraded performance) anunexpected power loss to one or more of the power supplies.

I think this is the setting we want on all chassis to ensure we use both power supplies under normal use and power-cap/throttle only when we would be exceeding the load supported (by 1 if the other is down, or both if both are available).

• Mode 2: Full AC/DC Redundancy ModePower-capping is enforced such that system has N+1 power redundancy. In this mode, the system willthrottle the nodes allowing for one of the available power supplies to be held in reserve. If only one powersupply is available, the system will throttle the nodes as to allow the full use of that power supply’scapacity.

This setting would make a chassis run off only one of the power supplies at all times. So that could simulate having only one power source (and seeing if the cartridges do fine with that).

Flags: needinfo?(q)

To be safe we should set a throttle mode.

Flags: needinfo?(q)

(In reply to Q from comment #3)

To be safe we should set a throttle mode.

THanks!

I set it to mode 1 on moon-chassis-1 and I'll set it on others once it works for a day or two without problems:

Edit chassis settings completed
Today 3:01:53 pm
    Auto power-on On
    Asset tag 05757
    Name moon-chassis-1
    Model HP Moonshot 1500 Chassis
    Power Mode Mode 1 - Max Performance with Redundancy
Flags: needinfo?(dhouse)

I set the power mode to 1 on all of the chassis yesterday.

Status: NEW → RESOLVED
Closed: 7 years ago
Flags: needinfo?(dhouse)
Resolution: --- → FIXED

No problems during the power outage for maintenance. there were some comments about the moonshots increased noise [to complain], but there were no chassis alerts that the power throttling happened (the chassis alerted that the power was dropped and then recovered, but no throttling).

Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.