- Resource: QueuedResource
- Tpu
- NodeSpec
- MultiNodeParams
- WorkloadType
- BestEffort
- Guaranteed
- Spot
- QueueingPolicy
- QueuedResourceState
- State
- CreatingData
- AcceptedData
- ProvisioningData
- FailedData
- DeletingData
- ActiveData
- SuspendingData
- SuspendedData
- StateInitiator
- ProvisioningModel
- RunDuration
- Methods
Resource: QueuedResource
A QueuedResource represents a request for resources that will be placed in a queue and fulfilled when the necessary resources are available.
| JSON representation |
|---|
{ "name" : string , "createTime" : string , "queueingPolicy" : { object ( |
name
string
Output only. Immutable. The name of the QueuedResource.
createTime
string (
Timestamp
format)
Output only. The time when the QueuedResource was created.
Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z"
, "2014-10-02T15:01:23.045123456Z"
or "2014-10-02T15:01:23+05:30"
.
queueingPolicy
object (
QueueingPolicy
)
The queueing policy of the QueuedRequest.
state
object (
QueuedResourceState
)
Output only. State of the QueuedResource request.
reservationName
string
Name of the reservation in which the resource should be provisioned. Format: projects/{project}/locations/{zone}/reservations/{reservation}
provisioningModel
enum (
ProvisioningModel
)
Optional. The provisioning model for the resource.
runDuration
object (
RunDuration
)
Optional. The duration of the requested resource.
resource
. Resource specification. resource
can be only one of the following:tpu
object (
Tpu
)
Defines a TPU resource.
tier
. Tier specifies the required tier. tier
can be only one of the following:bestEffort
object (
BestEffort
)
The BestEffort tier.
guaranteed
object (
Guaranteed
)
The Guaranteed tier.
spot
object (
Spot
)
Optional. The Spot tier.
Tpu
Details of the TPU resource(s) being requested.
| JSON representation |
|---|
{
"nodeSpec"
:
[
{
object (
|
| Fields | |
|---|---|
nodeSpec[]
|
The TPU node(s) being requested. |
NodeSpec
Details of the TPU node(s) being requested. Users can request either a single node or multiple nodes. NodeSpec provides the specification for node(s) to be created.
| JSON representation |
|---|
{ "parent" : string , "nodeId" : string , "multiNodeParams" : { object ( |
| Fields | |
|---|---|
parent
|
Required. The parent resource name. |
nodeId
|
The unqualified resource name. Should follow the |
multiNodeParams
|
Optional. Fields to specify in case of multi-node request. |
node
|
Required. The node. |
MultiNodeParams
Parameters to specify for multi-node QueuedResource requests. This field must be populated in case of multi-node requests instead of nodeId. It's an error to specify both nodeId and multiNodeParams.
| JSON representation |
|---|
{
"nodeCount"
:
integer
,
"nodeIdPrefix"
:
string
,
"workloadType"
:
enum (
|
| Fields | |
|---|---|
nodeCount
|
Required. Number of nodes with this spec. The system will attempt to provison "nodeCount" nodes as part of the request. This needs to be > 1. |
nodeIdPrefix
|
Prefix of node_ids in case of multi-node request Should follow the |
workloadType
|
Optional. The workload type for the multi-node request. |
WorkloadType
The workload type for the multi-node request.
| Enums | |
|---|---|
WORKLOAD_TYPE_UNSPECIFIED
|
Not specified. |
THROUGHPUT_OPTIMIZED
|
All of the nodes are available most of the time. Recommended for training workloads. |
AVAILABILITY_OPTIMIZED
|
Most of the nodes are available all of the time. Recommended for serving workloads. |
BestEffort
This type has no fields.
BestEffort tier definition.
Guaranteed
Guaranteed tier definition.
| JSON representation |
|---|
{ "minDuration" : string , "reserved" : boolean } |
| Fields | |
|---|---|
minDuration
|
Optional. Defines the minimum duration of the guarantee. If specified, the requested resources will only be provisioned if they can be allocated for at least the given duration. A duration in seconds with up to nine fractional digits, ending with ' |
reserved
|
Optional. Specifies the request should be scheduled on reserved capacity. |
Spot
This type has no fields.
Spot tier definition.
QueueingPolicy
Defines the policy of the QueuedRequest.
| JSON representation |
|---|
{ // Union field |
start_timing_constraints
. Time flexibility specification. start_timing_constraints
can be only one of the following:validUntilDuration
string (
Duration
format)
A relative time after which resources should not be created. If the request cannot be fulfilled by this time the request will be failed.
A duration in seconds with up to nine fractional digits, ending with ' s
'. Example: "3.5s"
.
validUntilTime
string (
Timestamp
format)
An absolute time after which resources should not be created. If the request cannot be fulfilled by this time the request will be failed.
Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z"
, "2014-10-02T15:01:23.045123456Z"
or "2014-10-02T15:01:23+05:30"
.
validAfterDuration
string (
Duration
format)
A relative time after which resources may be created.
A duration in seconds with up to nine fractional digits, ending with ' s
'. Example: "3.5s"
.
validAfterTime
string (
Timestamp
format)
An absolute time at which resources may be created.
Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z"
, "2014-10-02T15:01:23.045123456Z"
or "2014-10-02T15:01:23+05:30"
.
validInterval
object (
Interval
)
An absolute time interval within which resources may be created.
QueuedResourceState
QueuedResourceState defines the details of the QueuedResource request.
| JSON representation |
|---|
{ "state" : enum ( |
state
enum (
State
)
State of the QueuedResource request.
stateInitiator
enum (
StateInitiator
)
Output only. The initiator of the QueuedResources's current state. Used to indicate whether the SUSPENDING/SUSPENDED state was initiated by the user or the service.
state_data
. Further data for the state. state_data
can be only one of the following:creatingData
object (
CreatingData
)
Further data for the creating state.
acceptedData
object (
AcceptedData
)
Further data for the accepted state.
provisioningData
object (
ProvisioningData
)
Further data for the provisioning state.
failedData
object (
FailedData
)
Further data for the failed state.
deletingData
object (
DeletingData
)
Further data for the deleting state.
activeData
object (
ActiveData
)
Further data for the active state.
suspendingData
object (
SuspendingData
)
Further data for the suspending state.
suspendedData
object (
SuspendedData
)
Further data for the suspended state.
State
Output only state of the request
| Enums | |
|---|---|
STATE_UNSPECIFIED
|
State of the QueuedResource request is not known/set. |
CREATING
|
The QueuedResource request has been received. We're still working on determining if we will be able to honor this request. |
ACCEPTED
|
The QueuedResource request has passed initial validation/admission control and has been persisted in the queue. |
PROVISIONING
|
The QueuedResource request has been selected. The associated resources are currently being provisioned (or very soon will begin provisioning). |
FAILED
|
The request could not be completed. This may be due to some late-discovered problem with the request itself, or due to unavailability of resources within the constraints of the request (e.g., the 'valid until' start timing constraint expired). |
DELETING
|
The QueuedResource is being deleted. |
ACTIVE
|
The resources specified in the QueuedResource request have been provisioned and are ready for use by the end-user/consumer. |
SUSPENDING
|
The resources specified in the QueuedResource request are being deleted. This may have been initiated by the user, or the Cloud TPU service. Inspect the state data for more details. |
SUSPENDED
|
The resources specified in the QueuedResource request have been deleted. |
WAITING_FOR_RESOURCES
|
The QueuedResource request has passed initial validation and has been persisted in the queue. It will remain in this state until there are sufficient free resources to begin provisioning your request. Wait times will vary significantly depending on demand levels. When demand is high, not all requests can be immediately provisioned. If you need more reliable obtainability of TPUs consider purchasing a reservation. To put a limit on how long you are willing to wait, use timing constraints . |
CreatingData
This type has no fields.
Further data for the creating state.
AcceptedData
This type has no fields.
Further data for the accepted state.
ProvisioningData
This type has no fields.
Further data for the provisioning state.
FailedData
Further data for the failed state.
| JSON representation |
|---|
{
"error"
:
{
object (
|
| Fields | |
|---|---|
error
|
The error that caused the queued resource to enter the FAILED state. |
DeletingData
This type has no fields.
Further data for the deleting state.
ActiveData
This type has no fields.
Further data for the active state.
SuspendingData
This type has no fields.
Further data for the suspending state.
SuspendedData
This type has no fields.
Further data for the suspended state.
StateInitiator
The initiator of the QueuedResource's SUSPENDING/SUSPENDED state.
| Enums | |
|---|---|
STATE_INITIATOR_UNSPECIFIED
|
The state initiator is unspecified. |
USER
|
The current QueuedResource state was initiated by the user. |
SERVICE
|
The current QueuedResource state was initiated by the service. |
ProvisioningModel
Defines the provisioning model for the resource.
| Enums | |
|---|---|
PROVISIONING_MODEL_UNSPECIFIED
|
Provisioning model is unknown. |
STANDARD
|
Standard provisioning with user controlled runtime. |
SPOT
|
Spot provisioning with no guaranteed runtime. |
RESERVATION_BOUND
|
Reservation provisioning with runtime bound to the lifetime of the consumed reservation. |
FLEX_START
|
Provisioning with DWS Flex Start with max run duration. |
RunDuration
Defines the maximum lifetime of the requested resource.
| JSON representation |
|---|
{ // Union field |
duration
. The maximum lifetime of the requested resource. duration
can be only one of the following:maxRunDuration
string (
Duration
format)
The maximum duration of the requested resource.
A duration in seconds with up to nine fractional digits, ending with ' s
'. Example: "3.5s"
.
terminationTime
string (
Timestamp
format)
The time at which the requested resource will be terminated.
Uses RFC 3339, where generated output will always be Z-normalized and use 0, 3, 6 or 9 fractional digits. Offsets other than "Z" are also accepted. Examples: "2014-10-02T15:01:23Z"
, "2014-10-02T15:01:23.045123456Z"
or "2014-10-02T15:01:23+05:30"
.
Methods |
|
|---|---|
|
Creates a QueuedResource TPU instance. |
|
Deletes a QueuedResource TPU instance. |
|
Gets details of a queued resource. |
|
Gets the maintenance info for a queued resource. |
|
Lists queued resources. |
|
Perform manual maintenance on specific nodes of a QueuedResource. |
|
Resets a QueuedResource TPU instance |

