Skip to content

SYSTEM Cited by 1 source

Pyroscope (1.x)

Pyroscope 1.x was Grafana Labs' original open-source continuous-profiling database, built on the Cortex architecture — the same horizontally-scalable Prometheus-compatible foundation that Mimir and Loki also started from.

Why it's retained as a distinct page

The 1.x and 2.x lines are architecturally distinct enough that conflating them in one page loses signal. 1.x inherits Cortex's write-path-replication cost envelope; systems/pyroscope-2|2.0 retires it. This page canonicalises the 1.x architecture and the motivation for the rearchitecture.

1.x architecture (Cortex-derived)

Per Grafana Labs' own retrospective in the 2.0 launch post:

"The original Pyroscope architecture was based on Cortex, which is the same foundation that the Mimir and Loki projects started with. It worked, but it carried overhead that made large-scale continuous profiling expensive to run and operationally heavy."

The key Cortex-era traits:

  • Write-path replication across ingesters for durability — each profile write replicated N-way on the ingest side, doubling or tripling ingest-tier compute and network budget.
  • Coupled read and write paths — scaling one meant scaling the other.
  • Local storage on ingesters as the initial write target, with later flush to object storage.

These traits worked for Cortex's original use case (Prometheus metrics at moderate scale) but became expensive as profiling scaled up, especially because profiling payloads are larger than metrics datapoints and carry heavy symbolic information. (Source: sources/2026-04-22-grafana-introducing-pyroscope-2-0)

Successor

Pyroscope 2.0 — ground-up rearchitecture following the same Cortex-to-object-store pattern Mimir and Loki adopted. Eliminates write-path replication, decouples reads from writes, makes object storage the single source of truth. See patterns/observability-db-rearchitecture-cortex-to-object-store.

Seen in

Last updated · 517 distilled / 1,221 read