Yandex Managed Service for Prometheus® overview
Yandex Managed Service for Prometheus® is a monitoring system compatible with Prometheus. You can use it to collect, store, and query metrics from your containers, applications, and infrastructure. The system leverages the Prometheus data model and PromQL, enabling you to use your current Grafana dashboards.
Prometheus features multiple tools for collecting metrics, including libraries for widely-used programming languages and support for exporting metrics from external systems, such as databases, message queues, and more. Yandex Managed Service for Prometheus® enables you to use best practices of the Prometheus open-source community, while ensuring high scalability, availability, and security with no need to deploy Prometheus clusters and manage the internal monitoring infrastructure on your own. To get started with Yandex Managed Service for Prometheus®, install agents that support the Remote Write protocol.
Yandex Managed Service for Prometheus® is based on the system Yandex uses to monitor its internal services. The system automatically scales metric write, storage, and read operations depending on the workload. Monitoring data is replicated across two availability zones, ensuring high availability.
Getting started
-
Open the Yandex Monitoring home page and select a folder.
-
In the left-hand panel, select Prometheus.
-
Click Create workspace.
-
On the page that opens, you will see links to endpoints.
Use these endpoints to configure writing and reading the workspace metrics in Prometheus format.
Feature comparison
Note
The system features will be enhanced in upcoming releases.
| Feature | Prometheus | Yandex Managed Service for Prometheus® |
|---|---|---|
| Collecting metrics | Scrape, Pushgateway, Remote Write | Metric writes via Remote Write are supported. You can use Unified Agent from Yandex Cloud to send metrics from the virtual machine to Yandex Managed Service for Prometheus®. Any compatible metric collection agent will also be suitable, including Prometheus itself. |
| Long-term metric storage | Not intended for long-term metric storage, using third-party solutions | Support for long-term metric storage, with time-unlimited storage possible through decimation. |
| Reading metrics | Data and metadata reads via the HTTP API | Data and metadata reads via the HTTP API with some restrictions |
| Visualization | Expression browser, Grafana | Charts and dashboards in Monitoring. Support for Prometheus data source |
| Aggregation | Aggregation via recording rules | Support for existing recording rule files in YAML format, which you can upload and manage via the Monitoring UI and API. |
| Alerting | Alerting via alerting rules | Support for existing alerting rule files in YAML format, which you can upload and manage via the Monitoring UI and API. |
| Integration | Client libraries and exporters | Existing libraries and exporters. |
Current limitations
- The
NaNvalue is not supported and is treated as a missing point. - The
+Inf/-Infvalues may be processed incorrectly. - Alerting rules only support email and Telegram channels, with no dynamic routing.
staleness markers,exemplars, andnative histogramsare not supported.
Quotas and limits
Yandex Managed Service for Prometheus® has the following limits:
- Quotas are organizational constraints that can be changed by support on request.
- Limits are technical constraints of the Yandex Cloud architecture. You cannot change the limits.
If you need more resources, contact support at cloud@support.yandex.ru, detailing which quotas you want increased and by how much.
Quotas per endpoint
| Type of limit | Value |
|---|---|
| Maximum data write speed in Remote Write format | 1,000 requests/s and 80 MB/s |
| Number of metrics per write request | 10,000 |
| Number of Remote Read API read requests per second | 200 |
| Number of HTTP API read requests per second | 200 |
| Number of unique metrics | 20,000,000 |
| Maximum retention period for expired metrics 1 | 60 days |
1 Metrics getting no new values for 60 days are considered expired and removed. Metrics that keep getting new values are stored indefinitely.
Here is what you can do in Prometheus > Workspace > General information:
- View the number of written metrics.
- Go to the page with service dashboards and view the number of requests, errors, and HTTP status codes per second for various endpoints.
Limits per endpoint
| Type of limit | Value |
|---|---|
| Number of unique labels per metric | 29 |
The data is transferred as a time series defined by its metric and labels. In Yandex Managed Service for Prometheus®, one metric may contain up to 29 unique labels (including name), i.e., up to 28 custom labels.
There are no restrictions on the number of metric values; it is only limited by the basic Monitoring quota, i.e., the number of metrics per service in the folder. Each new combination of values is a new metric. For example, http_server_requests_seconds_count{uri="…"} may have many uri values, each creating a new time series and consuming the quota.
Once the quota is depleted, attempts to write to the new time series will fail with an error. Previously written data is not deleted. The quota limits the number of time series but not amount of data. The data will be deleted only if there are no records during the TTL, so you will be able to create new time series after you free up or increase your quota.
© 2025 Linux Foundation. All rights reserved. The Linux Foundation owns and uses registered trademarks. For a list of Linux Foundation trademarks, see Trademark Usage.