- Resource: ServiceLevelObjective
- ServiceLevelIndicator
- BasicSli
- AvailabilityCriteria
- LatencyCriteria
- RequestBasedSli
- TimeSeriesRatio
- DistributionCut
- Range
- WindowsBasedSli
- PerformanceThreshold
- MetricRange
- CalendarPeriod
- Methods
Resource: ServiceLevelObjective
A Service-Level Objective (SLO) describes a level of desired good service. It consists of a service-level indicator (SLI), a performance goal, and a period over which the objective is to be evaluated against that goal. The SLO can use SLIs defined in a number of different manners. Typical SLOs might include "99% of requests in each rolling week have latency below 200 milliseconds" or "99.5% of requests in each calendar month return successfully."
JSON representation |
---|
{ "name": string, "displayName": string, "serviceLevelIndicator": { object ( |
Fields | |
---|---|
name |
Identifier. Resource name for this
|
display |
Name used for UI elements listing this SLO. |
service |
The definition of good service, used to measure and calculate the quality of the |
goal |
The fraction of service that must be good in order for this objective to be met. |
user |
Labels which have been used to annotate the service-level objective. Label keys must start with a letter. Label keys and values may contain lowercase letters, numbers, underscores, and dashes. Label keys and values have a maximum length of 63 characters, and must be less than 128 bytes in size. Up to 64 label entries may be stored. For labels which do not have a semantic value, the empty string may be supplied for the label value. |
Union field period . The time period over which the objective will be evaluated. period can be only one of the following: |
|
rolling |
A rolling time period, semantically "in the past |
calendar |
A calendar period, semantically "since the start of the current |
ServiceLevelIndicator
A Service-Level Indicator (SLI) describes the "performance" of a service. For some services, the SLI is well-defined. In such cases, the SLI can be described easily by referencing the well-known SLI and providing the needed parameters. Alternatively, a "custom" SLI can be defined with a query to the underlying metric store. An SLI is defined to be good_service /
total_service
over any queried time interval. The value of performance always falls into the range 0 <= performance <= 1
. A custom SLI describes how to compute this ratio, whether this is by dividing values from a pair of time series, cutting a Distribution
into good and bad counts, or counting time windows in which the service complies with a criterion. For separation of concerns, a single Service-Level Indicator measures performance for only one aspect of service quality, such as fraction of successful queries or fast-enough queries.
JSON representation |
---|
{ // Union field |
Fields | |
---|---|
Union field type . Service level indicators can be grouped by whether the "unit" of service being measured is based on counts of good requests or on counts of good time windows type can be only one of the following: |
|
basic |
Basic SLI on a well-known service type. |
request |
Request-based SLIs |
windows |
Windows-based SLIs |
BasicSli
An SLI measuring performance on a well-known service type. Performance will be computed on the basis of pre-defined metrics. The type of the service_resource
determines the metrics to use and the service_resource.labels
and metricLabels
are used to construct a monitoring filter to filter that metric down to just the data relevant to this service.
JSON representation |
---|
{ "method": [ string ], "location": [ string ], "version": [ string ], // Union field |
Fields | |
---|---|
method[] |
OPTIONAL: The set of RPCs to which this SLI is relevant. Telemetry from other methods will not be used to calculate performance for this SLI. If omitted, this SLI applies to all the Service's methods. For service types that don't support breaking down by method, setting this field will result in an error. |
location[] |
OPTIONAL: The set of locations to which this SLI is relevant. Telemetry from other locations will not be used to calculate performance for this SLI. If omitted, this SLI applies to all locations in which the Service has activity. For service types that don't support breaking down by location, setting this field will result in an error. |
version[] |
OPTIONAL: The set of API versions to which this SLI is relevant. Telemetry from other API versions will not be used to calculate performance for this SLI. If omitted, this SLI applies to all API versions. For service types that don't support breaking down by version, setting this field will result in an error. |
Union field sli_criteria . This SLI can be evaluated on the basis of availability or latency. sli_criteria can be only one of the following: |
|
availability |
Good service is defined to be the count of requests made to this service that return successfully. |
latency |
Good service is defined to be the count of requests made to this service that are fast enough with respect to |
AvailabilityCriteria
This type has no fields.
Future parameters for the availability SLI.
LatencyCriteria
Parameters for a latency threshold SLI.
JSON representation |
---|
{ "threshold": string } |
Fields | |
---|---|
threshold |
Good service is defined to be the count of requests made to this service that return in no more than |
RequestBasedSli
Service Level Indicators for which atomic units of service are counted directly.
JSON representation |
---|
{ // Union field |
Fields | |
---|---|
Union field method . The means to compute a ratio of good_service to total_service . method can be only one of the following: |
|
good |
|
distribution |
|
TimeSeriesRatio
A TimeSeriesRatio
specifies two TimeSeries
to use for computing the good_service / total_service
ratio. The specified TimeSeries
must have ValueType = DOUBLE
or ValueType = INT64
and must have MetricKind =
DELTA
or MetricKind = CUMULATIVE
. The TimeSeriesRatio
must specify exactly two of good, bad, and total, and the relationship good_service +
bad_service = total_service
will be assumed.
JSON representation |
---|
{ "goodServiceFilter": string, "badServiceFilter": string, "totalServiceFilter": string } |
Fields | |
---|---|
good |
A monitoring filter specifying a |
bad |
A monitoring filter specifying a |
total |
A monitoring filter specifying a |
DistributionCut
A DistributionCut
defines a TimeSeries
and thresholds used for measuring good service and total service. The TimeSeries
must have ValueType =
DISTRIBUTION
and MetricKind = DELTA
or MetricKind = CUMULATIVE
. The computed good_service
will be the estimated count of values in the Distribution
that fall within the specified min
and max
.
JSON representation |
---|
{
"distributionFilter": string,
"range": {
object ( |
Fields | |
---|---|
distribution |
A monitoring filter specifying a |
range |
Range of values considered "good." For a one-sided range, set one bound to an infinite value. |
Range
Range of numerical values within min
and max
.
JSON representation |
---|
{ "min": number, "max": number } |
Fields | |
---|---|
min |
Range minimum. |
max |
Range maximum. |
WindowsBasedSli
A WindowsBasedSli
defines good_service
as the count of time windows for which the provided service was of good quality. Criteria for determining if service was good are embedded in the window_criterion
.
JSON representation |
---|
{ "windowPeriod": string, // Union field |
Fields | |
---|---|
window |
Duration over which window quality is evaluated. Must be an integer fraction of a day and at least |
Union field window_criterion . The criterion to use for evaluating window goodness. window_criterion can be only one of the following: |
|
good |
A monitoring filter specifying a |
good |
A window is good if its |
metric |
A window is good if the metric's value is in a good range, averaged across returned streams. |
metric |
A window is good if the metric's value is in a good range, summed across returned streams. |
PerformanceThreshold
A PerformanceThreshold
is used when each window is good when that window has a sufficiently high performance
.
JSON representation |
---|
{ "threshold": number, // Union field |
Fields | |
---|---|
threshold |
If window |
Union field type . The means, either a request-based SLI or a basic SLI, by which to compute performance over a window. type can be only one of the following: |
|
performance |
|
basic |
|
MetricRange
A MetricRange
is used when each window is good when the value x of a single TimeSeries
satisfies range.min <= x <= range.max
. The provided TimeSeries
must have ValueType = INT64
or ValueType = DOUBLE
and MetricKind = GAUGE
.
JSON representation |
---|
{
"timeSeries": string,
"range": {
object ( |
Fields | |
---|---|
time |
A monitoring filter specifying the |
range |
Range of values considered "good." For a one-sided range, set one bound to an infinite value. |
CalendarPeriod
A CalendarPeriod
represents the abstract concept of a time period that has a canonical start. Grammatically, "the start of the current CalendarPeriod
." All calendar times begin at midnight UTC.
Enums | |
---|---|
CALENDAR_PERIOD_UNSPECIFIED |
Undefined period, raises an error. |
DAY |
A day. |
WEEK |
A week. Weeks begin on Monday, following ISO 8601. |
FORTNIGHT |
A fortnight. The first calendar fortnight of the year begins at the start of week 1 according to ISO 8601. |
MONTH |
A month. |
QUARTER |
A quarter. Quarters start on dates 1-Jan, 1-Apr, 1-Jul, and 1-Oct of each year. |
HALF |
A half-year. Half-years start on dates 1-Jan and 1-Jul. |
YEAR |
A year. |
Methods |
|
---|---|
|
Create a ServiceLevelObjective for the given Service . |
|
Delete the given ServiceLevelObjective . |
|
Get a ServiceLevelObjective by name. |
|
List the ServiceLevelObjective s for the given Service . |
|
Update the given ServiceLevelObjective . |