Dasha 1.8 is the latest release of an open source performance dashboard for PostgreSQL fleets. It connects to clusters with a read-only role, reports what databases are doing right now, and suggests what to do about it. Nothing is installed on database hosts: no agent, and no extension beyond those already present. A single instance serves many clusters and all their replicas. Because most PostgreSQL statistics are per-instance and not replicated, Dasha reads every host of a cluster and reasons over the combined picture — an index that looks unused on the primary may be serving the entire read workload on a standby.
The concrete additions since the last announcement:
- Index recommendations: Dasha reads
pg_stat_statementson every host and proposes missing B-tree indexes, each with column order, a readyCREATE INDEX CONCURRENTLY, the statements it covers, and their share of execution time. - Write and partition awareness: write counters and partitioned tables (
ATTACH PARTITION) are taken into account when evaluating candidates. - Plan-backed evidence: candidates can be backed by
auto_explainplans from the log, showing sequential scans of the table, what they cost, and how many rows their filters discarded. - Statistics warnings: a misestimated row count raises a warning to run
ANALYZE. - Additional features: I/O analysis, schema checks, and log insights round out the release.
For developers, the practical value is that recommendations arrive with the evidence needed to act: the exact DDL, the queries affected, and the measured cost of the current sequential scans. The per-host reading model also guards against dropping indexes that only appear idle on the primary. Warnings about misestimated row counts point directly at missing ANALYZE runs, helping keep planner estimates accurate before tuning further.