May 2019
Intermediate to advanced
442 pages
11h 36m
English
Having multiple shards, or even just more than one datacenter, means that time series data is now spread out across different Prometheus instances. Hierarchical federation aims to solve this issue by having one or more Prometheus servers collecting high-level aggregate time series from other Prometheus instances. You can have more than two levels of federation, though that would require significant scale. This allows the higher-tiered Prometheus to have a broader view over the infrastructure and its applications. However, as only aggregated metrics should be federated, those with the most context and detail will still reside on the lower tiers. The following diagram illustrates how this works:
Read now
Unlock full access