This page describes the storage policies for data that is used by Cloud Monitoring. Cloud Monitoring is a global product and its services are available with no dependence on location.
Storage of time-series data
Google agents, client-side code using libraries like OpenCensus and OpenTelemetry, and third-party sources such as Blue Medora, write time series by using the Cloud Monitoring API. Google services also write time series, but they use internal mechanisms.
To store a time series in a specific region, the monitored resource against which the time series is written must have one of the following labels and the value of the label must be valid:
A time series is discarded when the monitored resource against which the time series is written has one of the previous labels and the value of the label isn't recognized or isn't specified.
The storage location of a time series is unspecified when the monitored resource against which the time series is written doesn't have one of the previous labels.
To view a list of all monitored resources and their labels, see Monitored resource types. For those monitored resources that contain a location label, the entry indicates the valid values for that label.
Data written by Google products
Google products deployable by zone or by region store any time series that they generate in the same region to which they are deployed. For a list of these products, see Products available by location.
For all other Google products, the storage location of any time series that the product generates is unspecified. For a list of these products, see Global products.
Time series for custom metrics follow the location rules described in Storage of time-series data. Examples of custom metrics include metrics you create from your applications by using libraries such as OpenCensus and OpenTelemetry.
Not all monitored resource types that are available for custom metrics
have a location label.
For example, the
global resource only has a label for
the project identifier. For a list of the monitored resource types available
for custom metrics,
see Monitored resources for custom metrics.
Processing of time-series data
The physical location where time-series data are processed when such data are queried depend on many factors including the storage location of the data, query semantics, networking, and others. In general, there is no guarantee that time series are always processed in the location where they are physically stored.
When Monitoring sends a notification of an alert, that information might pass through third-party providers that don't support data localization. For example, customers can configure an alert notification to be delivered to any email address, to a Slack channel, or through SMS. For details on all available channel types, see Notification options.