The Catalyst Cloud Metrics Service is free to use for all customers.
You will not be charged for the collection of metrics your cloud resources generate, or accessing them via the metrics API.
Sorry, we do not provide any dashboards that allow customers to visualise resource metrics for their projects at this time.
Unfortunately we do not support customers publishing their own metrics to the Metrics Service at this time.
With the exception of memory usage (see below), the Metrics Service does not use or require an agent to be installed on customer instances.
Metrics are collected external to workloads using non-invasive methods, and none of your personal data stored on the cloud is read or inspected by our telemetry services.
Memory usage metrics are exposed to our telemetry services using a device driver pre-installed into our operating system images.
These metrics are collected securely and anonymously on a hypervisor level. Only aggregated metrics such as total memory usage are collected; no additional information (such as the applications running on your system) are exposed or collected.
You can opt-out of this functionality by disabling
the virtio_balloon kernel driver in your instance.
Note
Memory usage metrics are not collected for Windows instances by default. Collecting memory usage metrics requires additional drivers to be installed.
For more information, see Enabling memory usage metrics for Windows instances.
With Compute C3-GPU and the other GPU-accelerated compute flavours, instances use PCI passthrough to access physical GPUs. This means that we are unable to collect metrics related to GPUs as they are under the full control of your compute instance.
We recommend collecting your own metrics using the NVIDIA DCGM Prometheus Exporter. If you’re using GPU acceleration on Catalyst Cloud Kubernetes Service, a Helm chart is available for installing the exporter into your cluster.
The following resource types currently do not have the “ended at” timestamp set when they are deleted:
VPNs
Load balancers
Database instances
Object storage containers
This will be implemented in the future, but in the meantime it is possible to infer that resources no longer exist through the absence of measures being published for their metrics.
On rare occasions the notifications that trigger the resource delete events might not be processed correctly. This means that every once in a while you might see that a deleted resource has not been marked as “ended” in the Metrics Service, even if it is not one of the above resource types. Applications using the Metrics Service should be implemented in such a way that it handles this case without any errors.
Once 90 days has passed after the resource is deleted, the Metrics Service will expire the corresponding resource metadata and metrics.
Please note that it is normal for a metric to not exist for newly created resources.
Metrics are only created for a resource if measures have been published for it; when the metric does not exist, that means that no measures have been published for it yet. Applications using the Metrics Service should be implemented with this in mind.
New metrics should be created with measures populated shortly.
Note
If the resource has existed for some time (more than 2 hours) and the resource or metric still does not exist in the Metrics Service, or exists but is not receiving new measures when it should, please raise a ticket via the Support Centre.