Graphite FAQ

Can I use tags?

Yes, our platform supports graphite tags.

How do I send data to the service?

See data ingestion.

Are there usage limits?

Grafana Cloud will reject metrics samples older than 1 hour with “out of bounds” error.

How does this compare to stock graphite?

Grafana Cloud uses the Graphite Proxy, which has some important differences with stock Graphite to be aware of:

Graphite Proxy is optimized for append-only workloads. Out of order writes are not supported by default, but you may reach out to support to have this feature enabled for you.
Graphite Proxy stores the raw data ingested and aggregates in runtime. Changes to schema files are applied retroactively. For more information, see Schema files and rollups.

My query results have a lower resolution than expected

There are two typical reasons why this happens: maxDataPoints and usage restrictions.

maxDataPoints sets a limit on the number of points returned per series, reducing resolution as necessary at runtime.

When querying through Grafana, maxDataPoints is set automatically. You can change it manually in the Query options. Refer to the query editor documentation for more information.
When querying directly through the HTTP API using cURL or similar, maxDataPoints is set to 800 by default. You can change this value by appending &maxDataPoints=<number> to your query.

Additionally, if a query tries to fetch more than 1 million datapoints, we will pick lower resolution data (rollups) for that query, to try to get the number under 1 million. If the query, no matter which rollup we try to choose, still tries to fetch more than 20 million datapoints, we reject it.

Example:

Let’s say you use the standard Grafana Cloud storage-schema rule: 10s:8d,10min:1y, and a query is issued for sumSeries(<metric name pattern>) and a time range of 24h. A metric with 10s resolution has 8640 data points in a 24h period, so if <metric name pattern> matches 120 or more metrics then 120*8640=1036800 which exceeds the threshold, so we will pick the 10m data instead. This way you typically still get high enough resolution (in this case 144 points per series), but higher query performance.

On custom plans, these limits can be adjusted. You can also pre-aggregate data in carbon-relay-ng to load fewer series. More information can be found in the Graphite querier documentation.

See also Graphite aggregation in Grafana Cloud for more details on how data is aggregated.

Do I have to exclusively use Grafana Cloud?

No, Grafana Cloud contains a data source that you can use however you like. E.g. in combination with other data sources, and queried from any Grafana instance or other client.