Executive Overview & Architectural Significance
Dasha 1.10 introduces profound architectural enhancements for open-source performance monitoring across large-scale PostgreSQL fleets. Operating entirely out-of-band without deploying local database agents or custom extensions beyond standard configurations, Dasha connects via minimal read-only roles to synthesize distributed cluster topologies. Because native PostgreSQL statistics remain per-instance and un-replicated, Dasha polls every cluster host to construct a unified telemetry picture. This multi-endpoint aggregation solves longstanding visibility blind spots—such as identifying indexes that appear inactive on a primary node yet actively serve heavy read workloads across standbys.
Version 1.10 expands high-availability infrastructure mapping by integrating directly with consensus stores (etcd v3 and Consul) used by Patroni. By reading Distributed Configuration Store (DCS) keys without modification, Dasha automatically translates Patroni scopes into comprehensive cluster maps. Members are indexed by address and name, system identifiers match primary-standby relationships even behind virtual IP architectures, and topology shifts propagate within the standard refresh interval. Furthermore, the inclusion of a read-only Model Context Protocol (MCP) server embeds fleet diagnostics directly into AI assistants, exposing 35 diagnostic tools and 5 specialized prompt workflows with per-user token passthrough.
Core Enhancements & Developer Ergonomics
Developer ergonomics receive a major boost via the new Activity & Waits dashboard for PostgreSQL 14+. This real-time interface plots active sessions grouped by CPU cycles, idle-in-transaction states, and specific wait event types (IO, Lock, LWLock) sampled every 1 to 10 seconds. Statement aggregation relies on native query_id calculation, decoupling analysis overhead from concurrent UI viewers. Additional telemetry features include predictive replication slot invalidation forecasting—measuring WAL generation rates via dual LSN server-clock samples without maintaining persistent connections—and precise progress tracking for COPY operations.
Health Score subsystems now support stock Prometheus and VictoriaMetrics scrapers (postgres_exporter, node_exporter, and pgbouncer_exporter), automatically mapping Patroni metrics by scope and member. Crucially, the scoring engine no longer penalizes cold tables that lack write activity past a configurable inactivity threshold, preventing idle schemas from degrading overall fleet health grades while offering targeted VACUUM (FREEZE, ANALYZE) maintenance cues.
Migration Notes & Upgrade Practicalities
Dasha 1.10 is fully backwards-compatible with existing deployments while requiring specific database capabilities for advanced features. Monitored instances must run PostgreSQL 14 through 19b, granting pg_monitor access and CONNECT privileges across target databases. Query statistics demand pg_stat_statements (or pgpro_stats), with compute_query_id enabled for real-time wait telemetry. Optional integrations like etcd, Consul, or centralized log engines require appropriate read permissions, and historical snapshot persistence benefits from a dedicated auxiliary PostgreSQL database.
