PCA Federation & Remote Storage Flashcards

What is 'cross-shard' federation typically used for in large-scale Prometheus deployments?

Answer: Merging metrics from multiple Prometheus shards that each scrape a subset of targets

When a single Prometheus cannot scrape all targets, multiple shards each cover a subset; a federation layer then pulls aggregated metrics from all shards into a unified view.

What does the 'write_relabel_configs' option inside a remote_write block allow you to do?

Answer: Filter or transform metric series before they are sent to the remote storage endpoint

write_relabel_configs applies relabeling rules to metric series just before remote_write sends them, allowing operators to drop, modify, or filter metrics destined for remote storage.

What is the main advantage of Cortex or Thanos over a standalone Prometheus for production long-term storage?

Answer: They provide horizontally scalable, multi-tenant storage with global query capability across instances

Cortex and Thanos both decouple storage from individual Prometheus instances, enabling scalable multi-tenant long-term retention and unified querying across many Prometheus servers.

When configuring the match[] parameter on the /federate endpoint, which selector would correctly fetch all metrics from the 'node' job?

Answer: {job="node"}

The match[] parameter accepts standard PromQL instant vector selectors using curly-brace syntax, so {job="node"} is the correct form to select all series for the node job.

What are 'exemplars' in Prometheus and how do they relate to remote storage?

Answer: Trace ID references attached to metric observations that can be forwarded via remote_write

Exemplars are optional references to distributed trace IDs stored alongside histogram or summary observations; they can be forwarded through remote_write to link metrics with traces in a compatible backend.

What is the purpose of a 'stale marker' sent by Prometheus via remote_write?

Answer: Signaling to the remote storage backend that a time series has stopped being scraped

When Prometheus detects that a target is no longer producing a metric (e.g., the target disappeared), it sends a special NaN stale marker so remote storage can properly close that series.

In a Prometheus remote_write configuration, what does the 'url' field specify?

Answer: The HTTP endpoint of the remote storage system that will receive the written metric samples

The url field in remote_write defines the remote storage endpoint (such as a Thanos receive component or Cortex distributor) where Prometheus will POST metric samples.