pgEdge/radar v0.5.0

on GitHub

Added

PostgreSQL

• stat_ssl collector - per-backend SSL/TLS state
• stat_replication_slots collector - replication slot spill counters (PG14+)

Per-database

• sequences collector - sequences with last/min/max values (PG10+)
• bloat collector - heuristic table bloat estimate from pg_stats
• pgstattuple collector - authoritative bloat via pgstattuple_approx(); opt-in with -include pgstattuple

CLI

• -exclude name1,name2 flag - skip named tasks
• -include name1,name2 flag - enable default-disabled tasks

Changed

• databases collector adds datfrozenxid, datminmxid, datconnlimit, datistemplate, datallowconn
• tables collector rebuilt on pg_class + pg_stat_all_tables; adds dead-tup counters, vacuum/analyze timestamps + ages, reloptions, relpersistence, relpages, heap/table/toast sizes; capped at top 1000 by size
• indexes collector rebuilt on pg_index + pg_stat_all_indexes; adds semantic-key columns, indisvalid, idx_scan/idx_tup_read/idx_tup_fetch, relpages, index_size; capped at top 1000 by size
• running_activity_maxage collector adds max_lock_wait_age
• available_extensions collector now sourced from pg_available_extension_versions (one row per version with trust/requires/schema metadata)
• Archive root now contains radar.out with the radar binary's version (build-time stamp) and commit

Fixed

• read_ahead collector ignored virtio/xen block devices and produced unlabeled output
• hostname.out reported resolver-canonical FQDN instead of kernel hostname; added hostname_fqdn.out for the FQDN
• pg_table_size/pg_relation_size called for every relation pre-sort caused fstat overload on >1000-table clusters; now uses relpages × block_size estimate with fstat fallback for relpages = 0 rows