Computer System Reset action schema definition conflict

AMI_Mani
Tuna

AMI

Posts: 175

Computer System Reset action schema definition conflict Mar 24, 2022 18:15:10 GMT

Quote

Post by AMI_Mani on Mar 24, 2022 18:15:10 GMT

Hi,
I'm able to see below content in ComputerSystem.v1_17_0.json for ForceRestart(like power cycle)

"Reset": {
"additionalProperties": false,
"description": "This action resets the system.",
"longDescription": "This action shall reset the system represented by the resource. For systems that implement ACPI Power Button functionality, the PushPowerButton value shall perform or emulate an ACPI Power Button Push. The ForceOff value shall remove power from the system or perform an ACPI Power Button Override, commonly known as a four-second hold of the Power Button. The ForceRestart value shall perform a ForceOff action, followed by an On action.",
"parameters": {
"ResetType": {
"$ref": "http://redfish.dmtf.org/schemas/v1/Resource.json#/definitions/ResetType",
"description": "The type of reset.",
"longDescription": "This parameter shall contain the type of reset. The service can accept a request without the parameter and perform an implementation-specific default reset. Services should include the @redfish.AllowableValues annotation for this parameter to ensure compatibility with clients, even when ActionInfo has been implemented."
}
},

But in redfish.dmtf.org/schemas/v1/Resource.json#/definitions/ResetType, Forcerestart is not like power cycle operation from definition

"ResetType": {

"enum": [

"On",

"ForceOff",

"GracefulShutdown",

"GracefulRestart",

"ForceRestart",

"Nmi",

"ForceOn",

"PushPowerButton",

"PowerCycle",

"Suspend",

"Pause",

"Resume"

],

"enumDescriptions": {

"ForceOff": "Turn off the unit immediately (non-graceful shutdown).",

"ForceOn": "Turn on the unit immediately.",

"ForceRestart": "Shut down immediately and non-gracefully and restart the system.",

"GracefulRestart": "Shut down gracefully and restart the system.",

"GracefulShutdown": "Shut down gracefully and power off.",

"Nmi": "Generate a diagnostic interrupt, which is usually an NMI on x86 systems, to stop normal operations, complete diagnostic actions, and, typically, halt the system.",

"On": "Turn on the unit.",

"Pause": "Pause execution on the unit but do not remove power. This is typically a feature of virtual machine hypervisors.",

"PowerCycle": "Power cycle the unit. Behaves like a full power removal, followed by a power restore to the resource.",

"PushPowerButton": "Simulate the pressing of the physical power button on this unit.",

"Resume": "Resume execution on the paused unit. This is typically a feature of virtual machine hypervisors.",

"Suspend": "Write the state of the unit to disk before powering off. This allows for the state to be restored when powered back on."

},

ForceRestart definition is different between two places, please confirm whether it's expected or it will be changed in schema?

Thanks,
Mani

jautor
Administrator

Posts: 502

Computer System Reset action schema definition conflict Mar 25, 2022 15:57:43 GMT

Quote

Post by jautor on Mar 25, 2022 15:57:43 GMT

Over time, as support for more types of equipment were added to Redfish, we created the common (in Resource schema) "ResetType" definition, which avoids terminology and references to ACPI as that applies only to computer/server products. But we didn't want to lose those computer-specific definitions, so they were kept in the ComputerSystem's Reset description.

Reading that description now, I'll suggest that we need to combine the two sentences that describe ACPI-based behavior, and remove the ForceRestart sentence, which as you point out, now describes only one possible implementation. The description of ForceRestart in Resource (and therefore the enumerations in ComputerSystem) is the better one.

Jeff

AMI_Mani
Tuna

AMI

Posts: 175

Computer System Reset action schema definition conflict Mar 25, 2022 17:47:58 GMT

Quote

Post by AMI_Mani on Mar 25, 2022 17:47:58 GMT

Hi Jeff,
Thanks for the reply.

in reset type (http://redfish.dmtf.org/schemas/v1/Resource.json#/definitions/ResetType)

enumDescriptions it's like below

"ForceRestart": "Shut down immediately and non-gracefully and restart the system.",

In enumLongDescriptions it's like below

ForceRestart": "This value shall indicate the resource will transition to a power on state, after transiting through a restart. The transition will start immediately. Upon successful completion, the PowerState property, if supported, shall shall contain the value `On`.",

Can you please explain about will transition to a power on state? What is difference between force restart and power cycle?

Thanks,
Mani

jautor
Administrator

Posts: 502

Computer System Reset action schema definition conflict Mar 30, 2022 1:10:23 GMT

Quote

Post by jautor on Mar 30, 2022 1:10:23 GMT

Mar 25, 2022 17:47:58 GMT AMI_Mani said:

Can you please explain about will transition to a power on state? What is difference between force restart and power cycle?

Thanks,
Mani

In theory, a "ForceRestart" means an immediate shutdown followed by a reboot/restart of the operating system - but does not require the power to be removed from the system. However, many designs will remove power when performing a reboot. The "PowerCycle" requires that power be removed from the unit/device and then restored - and doesn't make any statements about how the system is shut down.

Because of the multiple ways a computer system can be reset / restarted, and because Redfish is intended to support "non-computer" devices, there is overlap in many of these reset type definitions. I would recommend that you support as many of the ResetTypes as possible - and just know that based on your implementation in hardware, some of them may simply be duplicates. For example, if your design always performs a power cycle during a reboot, the "ForceRestart" and "PowerCycle" values would perform the same actions and have the same result.

And the somewhat clumsy language of "will transition to a power on state" is just trying to say that regardless of the state of the unit prior to the Reset action, at the end of the process, the unit should be powered on. We had some early implementations that would not power on if a "ForceRestart" was issued while the power was off. That's not how RESTful interfaces are supposed to work - you need to get to the desired end state regardless of your starting state...

Hope that helps,

Jeff

AMI_Mani
Tuna

AMI

Posts: 175

Computer System Reset action schema definition conflict Jul 27, 2022 19:09:53 GMT

Quote

Post by AMI_Mani on Jul 27, 2022 19:09:53 GMT

Thanks Jeff for the details.
Redfish service implementation can support some of ResetType enumeration based on hardware and it's not required to implement all. Do we have any information in specification for supporting some of ResetType enumeration only. I'm able to see this information related to allowable values

To specify the list of supported values for a parameter, the service may include the

@redfish.AllowableValues annotation.

Do we have any information like which ResetType enumeration values are mandatory or we can implement only some values from ResetType enumeration based on Hardware?

Thanks,
Mani

mraineri
Administrator

Posts: 1,048

Computer System Reset action schema definition conflict Jul 28, 2022 15:30:03 GMT

Quote

Post by mraineri on Jul 28, 2022 15:30:03 GMT

Redfish does not make any of the values mandatory. However, different organizations and customers consuming Redfish, like OCP, may have requirements for mandatory values to support for their environments.

As a starting point, it would be best to do what makes the most sense for your product and underlying architecture.

jautor
Administrator

Posts: 502

Computer System Reset action schema definition conflict Jul 29, 2022 17:57:05 GMT

Quote

Post by jautor on Jul 29, 2022 17:57:05 GMT

I would add that you should also list support for the ResetTypes that are functionally equivalent for your product, to give you the best chance at interoperability. Because there are so many "flavors" of Reset, it can be difficult for a client to find one type that works across the board - but supporting as many as you can, you raise the odds that a generic script will work on your product without modification.

For example, if your product doesn't distinguish between a 'graceful restart', a 'forced restart` and a `power cycle`, supporting all three of those ResetTypes (all performing the same process) makes it easier for a client to "find something in common" with other implementations.

Jeff