Default Thresholds Reference

This page lists every built-in check a fresh install seeds, together with the default threshold it ships with — the exact rules pghf.evaluate_run() judges each result against until you override them. It is generated from a fully-seeded catalog: 175 checks, 114 default thresholds.

To see the live values on your own install (which may differ if you've called pghf.set_threshold() since installing):

SELECT * FROM pghf.list_thresholds();

How to read the tables

• Judged as is the check's classification, fixed in the catalog: lower is better means a tier breaches when the value exceeds it, higher is better when the value falls below it (so for higher-is-better checks the smallest cutoff is the most severe tier). Boolean checks show their pass value; the threshold names the severity assigned when the observed value doesn't match it.
• A rule of > 0 on a count means "any at all" — a single occurrence fires that tier.
• Threshold values are in the check's native unit, shown in the Result column (kilobytes, pages, milliseconds, …).
• none — … rows are deliberate: the stated reason is recorded as a comment in sql/24_seed_data.sql alongside the seeding calls. A check with no threshold (and no custom evaluator) evaluates as not_evaluable, which ranks below ok and never alerts.
• none yet — … rows are text/jsonb-typed checks the default comparator cannot judge; each needs a custom evaluator_routine (seven built-ins already have one — see Built-in Custom Evaluators).
• Breach confirmation is instant for every seeded default: no check ships with breach-side debounce (min_consecutive_breaches stays at 1) — a check's very first breaching run always confirms right away, so a fresh install genuinely reveals real problems on its first run. Recovery is a different story — see the note below.

Seeded hysteresis: two confirmations to clear, not one

Every check above that has a threshold (114 of them, minus the 11 seeded-on-demand/category-10 ones — see Suites) also ships with min_consecutive_clears => 2 and a min_consecutive_window sized to twice that check's own suite cadence — clearing a confirmed breach needs 2 consecutive clean evaluated runs, not just one. This is a count of runs, not a minimum elapsed wall-clock time: if you've just fixed the underlying problem, calling pghf.run_and_evaluate() twice in a row clears it immediately — you don't need to wait for the check's normal schedule. What this actually protects against is a single lucky/noisy sample clearing a still-real problem on an unattended scheduled run, not a DBA who just fixed and manually re-verified it. This is orthogonal to the tiers in the tables below; see Symmetric recovery for the mechanics and the exact per-cadence window sizes. It's also only a recommendation calibrated to a check's assigned suite name — this project has no scheduler, so nothing guarantees a check actually runs that often; retune with pghf.set_threshold() if your real usage differs.

Prerequisite-gated checks

Five checks register only when their backing extension is installed on the node at seed time: PGHF03-012, PGHF04-008, PGHF04-011 (pg_stat_statements), PGHF07-007 (amcheck), and PGHF08-006 (pg_visibility). PGHF04-011's threshold below is likewise only seeded when that check registered. On a node without those extensions you'll see 170 checks and 113 thresholds instead.

Every threshold here is a starting point, not a verdict — update any of them for your site with pghf.set_threshold() (one threshold per check, applying wherever the check runs). For what each check measures and how to fix a breach, follow the category headings into the Check Reference pages.

C01 — Connection & Availability

Check	Name	Result	Judged as	Default threshold
PGHF01-001	SSL/TLS enabled	boolean	pass = true	warning when false
PGHF01-002	PostgreSQL version	integer (years_past_eol)	lower is better	critical > 1 · warning > 0 · info > -1
PGHF01-003	Connection saturation	numeric (percent)	lower is better	critical > 90 · warning > 75
PGHF01-004	Oldest idle-in-transaction session age	integer (seconds)	lower is better	warning > 30
PGHF01-005	Per-database connection counts	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF01-006	pg_hba TRUST on non-loopback	integer (count)	lower is better	critical > 0

C02 — pgBackRest Configuration & WAL Archiving

Check	Name	Result	Judged as	Default threshold
PGHF02-001	archive_command	text	—	none yet — text result needs a custom evaluator
PGHF02-002	pg_stat_archiver failures	integer (count)	lower is better	warning > 1
PGHF02-003	Incremental-backup readiness (summarize_wal)	boolean	pass = true	warning when false

C03 — Performance Parameters

Check	Name	Result	Judged as	Default threshold
PGHF03-001	shared_buffers eviction ratio	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF03-002	work_mem worst-case total	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF03-003	maintenance_work_mem	integer (kilobytes)	higher is better	critical < 131072 · warning < 163840 · info < 262144
PGHF03-004	effective_cache_size	integer (pages)	higher is better	info < 524289
PGHF03-005	max_parallel_workers_per_gather	integer (workers)	higher is better	warning < 2
PGHF03-006	min_parallel_table_scan_size	integer (pages)	lower is better	warning > 131072 · info > 1024
PGHF03-007	checkpoint_completion_target	numeric (fraction)	higher is better	critical < 0.5 · warning < 0.7 · info < 0.9
PGHF03-008	Requested checkpoint ratio	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF03-009	wal_compression	text	—	none yet — text result needs a custom evaluator
PGHF03-010	random_page_cost	numeric (relative_cost)	lower is better	warning > 4.0 · info > 1.1
PGHF03-011	effective_io_concurrency	integer (count)	higher is better	info < 2
PGHF03-012	JIT overhead	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF03-013	wal_buffers	integer (pages)	higher is better	info < 512
PGHF03-014	default_statistics_target	integer (count)	higher is better	info < 100
PGHF03-015	Temp file spill	integer (bytes)	lower is better	critical > 10737418240
PGHF03-016	Database cache hit ratio	numeric (percent)	higher is better	warning < 90 · info < 95
PGHF03-017	track_io_timing	text	—	none yet — text result needs a custom evaluator
PGHF03-018	Parallel worker pool saturation	numeric (percent)	lower is better	critical > 95 · warning > 85 · info > 70
PGHF03-019	track_wal_io_timing	boolean	pass = true	warning when false
PGHF03-020	pg_stat_io relation-extend pressure	numeric (milliseconds)	lower is better	none — cumulative-since-reset counter; a flat cutoff would eventually fire regardless of current health; it needs a baseline evaluator

C04 — Long-Running Queries & Lock Contention

Check	Name	Result	Judged as	Default threshold
PGHF04-001	Longest-running query age	integer (seconds)	lower is better	critical > 300 · warning > 60
PGHF04-002	Idle-in-transaction age	integer (seconds)	lower is better	critical > 300
PGHF04-003	Lock blocker chains	integer (count)	lower is better	critical > 20 · warning > 5 · info > 0
PGHF04-004	Deadlock count	integer (count)	lower is better	warning > 0
PGHF04-005	statement_timeout	integer (milliseconds)	higher is better	info < 1
PGHF04-006	idle_in_transaction_session_timeout	integer (milliseconds)	higher is better	warning < 1
PGHF04-007	pg_stat_statements extension	boolean	pass = true	warning when false
PGHF04-008	Top queries by total_exec_time	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF04-009	log_min_duration_statement	integer (milliseconds)	higher is better	info < 0
PGHF04-010	lock_timeout	integer (milliseconds)	higher is better	info < 1
PGHF04-011	Worst mean exec time (statistically-significant queries)	numeric (milliseconds)	lower is better	warning > 5000

C05 — Vacuum & Autovacuum Health

Check	Name	Result	Judged as	Default threshold
PGHF05-001	Database TXID age	numeric (transactions)	lower is better	critical > 1800000000 · warning > 1000000000
PGHF05-002	Table TXID age	numeric (transactions)	lower is better	critical > 1800000000 · warning > 1000000000
PGHF05-003	Dead tuple ratio	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF05-004	Last autovacuum age	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF05-005	Tables with autovacuum disabled	integer (count)	lower is better	warning > 0
PGHF05-006	autovacuum_vacuum_scale_factor	numeric (fraction)	lower is better	warning > 0.4 · info > 0.2
PGHF05-007	autovacuum_max_workers	integer (count)	higher is better	info < 3
PGHF05-008	autovacuum_vacuum_cost_delay	numeric (milliseconds)	lower is better	warning > 10 · info > 2
PGHF05-009	Table bloat estimate	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF05-010	In-progress vacuum age	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF05-011	autovacuum_work_mem	integer (kilobytes)	higher is better	info < 0
PGHF05-012	Multixact wraparound age	numeric (multixact_ids)	lower is better	critical > 1800000000 · warning > 1000000000
PGHF05-013	Oldest prepared transaction age	integer (seconds)	lower is better	warning > 3600
PGHF05-014	Autovacuum worker saturation	numeric (percent)	lower is better	critical > 98 · warning > 90 · info > 80
PGHF05-015	Table statistics staleness	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF05-016	pg_stat_io eviction pressure by context	jsonb	—	none — what counts as "too much" eviction scales with shared_buffers and workload; needs a baseline evaluator

C06 — Index Health

Check	Name	Result	Judged as	Default threshold
PGHF06-001	Unused indexes	integer (count)	lower is better	warning > 10 · info > 0
PGHF06-002	Duplicate indexes	integer (count)	lower is better	warning > 5 · info > 0
PGHF06-003	Invalid indexes	integer (count)	lower is better	critical > 3 · warning > 0
PGHF06-004	Index size vs row count	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF06-005	FK columns without index	integer (count)	lower is better	warning > 10 · info > 0
PGHF06-006	Prefix-redundant indexes	integer (count)	lower is better	warning > 5 · info > 0
PGHF06-007	Low-cardinality indexed columns	integer (count)	lower is better	info > 0
PGHF06-008	BRIN index correlation	integer (count)	lower is better	info > 0
PGHF06-009	Statistics reset date	text	—	none yet — text result needs a custom evaluator
PGHF06-010	Tables without primary key	integer (count)	lower is better	warning > 10 · info > 0

C07 — TOAST Table & Corruption Detection

Check	Name	Result	Judged as	Default threshold
PGHF07-001	data_checksums	boolean	pass = true	warning when false
PGHF07-002	Checksum failures	integer (count)	lower is better	critical > 0
PGHF07-003	TOAST reference integrity	integer (count)	lower is better	critical > 0
PGHF07-005	Orphaned TOAST tables	integer (count)	lower is better	critical > 3 · warning > 0
PGHF07-006	TOAST size vs main table	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF07-007	amcheck index verification	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF07-008	Cache hit ratio (user tables)	numeric (percent)	higher is better	warning < 90 · info < 95
PGHF07-009	Relation integrity probe availability	boolean	pass = true	info when false
PGHF07-010	default_toast_compression	text	—	custom evaluator — pghf.evaluate_c07_010_toast_compression()

C08 — Visibility Map Integrity

Check	Name	Result	Judged as	Default threshold
PGHF08-001	Heap block reads vs index scans	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF08-002	relallvisible consistency	integer (count)	lower is better	warning > 20 · info > 0
PGHF08-003	Post-crash visibility map advisory	boolean	—	none — the value is pg_is_in_recovery() — context, not pass/fail (a standby would breach forever)
PGHF08-004	pg_visibility extension	boolean	pass = true	info when false
PGHF08-005	Dead tuple counter anomaly candidates	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF08-006	VM integrity (pg_visibility)	jsonb	—	none yet — jsonb result needs a custom evaluator

C09 — WAL & Replication Slot Health

Check	Name	Result	Judged as	Default threshold
PGHF09-001	Worst inactive replication slot lag	integer (bytes)	lower is better	critical > 524288000 · warning > 52428800
PGHF09-002	Worst retained WAL vs max_slot_wal_keep_size	numeric (percent)	lower is better	critical > 90 · warning > 70
PGHF09-003	Replication slot count vs max_replication_slots	numeric (percent)	lower is better	warning > 80
PGHF09-004	Replication lag (pg_stat_replication)	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF09-005	Unnamed standbys	integer (count)	lower is better	info > 0
PGHF09-006	recovery_min_apply_delay	text	—	none yet — text result needs a custom evaluator
PGHF09-007	wal_keep_size vs unslotted standbys	integer (count)	lower is better	warning > 0
PGHF09-008	WAL archiving cross-reference	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF09-009	Invalidated replication slots	integer (count)	lower is better	warning > 0
PGHF09-010	max_slot_wal_keep_size	integer (megabytes)	higher is better	warning < 0
PGHF09-011	Inactive logical slots	integer (count)	lower is better	critical > 0
PGHF09-012	Logical replication subscription health	jsonb	—	custom evaluator — pghf.evaluate_c09_012_subscription_health()
PGHF09-013	Subscription table sync state	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF09-014	Streaming replication lag (time)	jsonb	—	custom evaluator — pghf.evaluate_c09_014_replication_lag()
PGHF09-015	logical_decoding_work_mem	integer (kilobytes)	higher is better	info < 65536
PGHF09-016	Failover-ready logical slot synchronization	integer (count)	lower is better	warning > 0

C10 — pg_upgrade Readiness

Check	Name	Result	Judged as	Default threshold
PGHF10-001	Legacy extensions (tsearch2/plpython2u)	integer (count)	lower is better	warning > 0
PGHF10-002	PostGIS version check	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF10-003	abstime/reltime/tinterval columns	integer (count)	lower is better	warning > 0
PGHF10-004	money type columns	integer (count)	lower is better	info > 0
PGHF10-005	SQL_ASCII databases	integer (count)	lower is better	warning > 0
PGHF10-006	Collation version	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF10-007	Tables approaching 1600-column limit	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF10-008	Pending prepared transactions	integer (count)	lower is better	warning > 0
PGHF10-009	Logical slots on upgrade	integer (count)	lower is better	warning > 0
PGHF10-010	C-language functions	integer (count)	lower is better	info > 0
PGHF10-011	plpython2u functions	integer (count)	lower is better	warning > 0
PGHF10-012	Ghost extensions	integer (count)	lower is better	warning > 0
PGHF10-013	Large objects	integer (count)	lower is better	warning > 1000000 · info > 100000
PGHF10-014	Custom tablespace paths	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF10-015	datconnlimit=0 databases	integer (count)	lower is better	warning > 0

C11 — Security Posture

Check	Name	Result	Judged as	Default threshold
PGHF11-001	Superusers without password	integer (count)	lower is better	warning > 0
PGHF11-002	MD5 authentication entries	integer (count)	lower is better	warning > 0
PGHF11-003	Public schema CREATE privilege	boolean	pass = false	warning when true
PGHF11-004	Privileged non-superuser roles	integer (count)	—	none — informational audit list for human review
PGHF11-005	Connection logging	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF11-006	pgaudit extension	boolean	pass = true	info when false
PGHF11-007	Non-superuser login role count	integer (count)	—	none — informational audit list for human review
PGHF11-008	SSL certificate paths	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF11-009	Superuser login roles as % of all login roles	numeric (percent)	lower is better	critical > 50 · warning > 20
PGHF11-010	password_encryption is scram-sha-256	boolean	pass = true	warning when false
PGHF11-011	Unencrypted client connections	integer (count)	lower is better	warning > 0
PGHF11-012	MAINTAIN privilege / pg_maintain role audit	integer (count)	—	none — informational audit list for human review

C12 — pgEdge / Spock Cluster

Check	Name	Result	Judged as	Default threshold
PGHF12-001	Node list consistency	boolean	pass = true	critical when false
PGHF12-002	Spock subscriptions enabled	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-003	Spock worker status	integer (count)	higher is better	warning < 1
PGHF12-004	Spock apply lag	jsonb	—	custom evaluator — pghf.evaluate_c12_004_apply_lag()
PGHF12-005	Spock exception log	integer (count)	lower is better	critical > 100000 · warning > 10000
PGHF12-006	Spock resolutions	integer (count)	—	none — a high count just means conflict resolution is firing; not necessarily a problem
PGHF12-007	Oldest unresolved spock exception age	numeric (days)	lower is better	warning > 7
PGHF12-008	pg_cron extension	boolean	—	none — presence is a deployment choice; no universal pass value
PGHF12-009	Spock WAL slots	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-010	hot_standby_feedback	text	—	none yet — text result needs a custom evaluator
PGHF12-011	wal_level	text	—	none yet — text result needs a custom evaluator
PGHF12-012	Table parity	boolean	pass = true	critical when false
PGHF12-013	Index parity	boolean	pass = true	warning when false
PGHF12-014	Sequence increment collision	boolean	pass = false	critical when true
PGHF12-015	Spock forward_origins	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-016	Replication set membership	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-017	Spock sync state	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-018	Row count sampling	boolean	pass = true	warning when false
PGHF12-019	Spock local node registration	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-020	Spock lag tracker	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF12-021	Spock queue depth	integer (count)	lower is better	custom evaluator — pghf.evaluate_c12_021_queue_depth()
PGHF12-022	Total spock channel conflicts	integer (count)	lower is better	warning > 1000
PGHF12-023	Worst spock replication progress lag	integer (bytes)	lower is better	critical > 524288000
PGHF12-024	Worst logical slot spill size	integer (bytes)	lower is better	critical > 2097152000 · warning > 524288000
PGHF12-025	Oldest unresolved spock exception age (retry/crash-loop)	integer (minutes)	lower is better	warning > 60

C13 — OS & Resource-Level Checks

Check	Name	Result	Judged as	Default threshold
PGHF13-001	Checkpoint sync time	numeric (milliseconds)	lower is better	none — cumulative-since-reset counter; a flat cutoff would eventually fire regardless of current health; it needs a baseline evaluator
PGHF13-002	pg_stat_io evictions	integer (count)	lower is better	none — cumulative-since-reset counter; a flat cutoff would eventually fire regardless of current health; it needs a baseline evaluator
PGHF13-003	maxwritten_clean	integer (count)	lower is better	none — cumulative-since-reset counter; a flat cutoff would eventually fire regardless of current health; it needs a baseline evaluator
PGHF13-004	huge_pages setting	text	—	none yet — text result needs a custom evaluator
PGHF13-005	Temp file spill	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF13-006	Query conflicts	integer (count)	lower is better	none — cumulative-since-reset counter; a flat cutoff would eventually fire regardless of current health; it needs a baseline evaluator
PGHF13-007	max_connections advisory	integer (count)	lower is better	warning > 500 · info > 200
PGHF13-008	Postmaster uptime	integer (seconds)	higher is better	critical < 300 · warning < 900 · info < 3600
PGHF13-009	Standby restartpoint completion	numeric (percent)	higher is better	warning < 90
PGHF13-010	Async I/O method	text	—	none — informational only — no claim about which io_method is correct for a given platform

C14 — WAL Growth & Generation Rate

Check	Name	Result	Judged as	Default threshold
PGHF14-001	pg_wal directory size	integer (bytes)	lower is better	critical > 53687091200 · warning > 21474836480
PGHF14-002	WAL generation rate	numeric (bytes/second)	lower is better	custom evaluator — pghf.evaluate_c14_002_wal_rate_baseline()
PGHF14-003	WAL statistics summary	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF14-004	Full-page write ratio	numeric (percent)	lower is better	warning > 40
PGHF14-005	Top WAL-generating tables	integer (count)	—	none — the scalar is a population count; the real signal is the top-N detail list
PGHF14-006	WAL compression	text	—	none yet — text result needs a custom evaluator
PGHF14-007	wal_level vs logical replication usage	jsonb	—	none yet — jsonb result needs a custom evaluator
PGHF14-008	WAL segment count	integer (count)	lower is better	warning > 1000
PGHF14-009	WAL archiver status	jsonb	—	custom evaluator — pghf.evaluate_c14_010_archiver_status()
PGHF14-010	UNLOGGED tables advisory	integer (count)	—	none — UNLOGGED tables are often an intentional choice
PGHF14-011	Forced checkpoint rate	integer (count)	lower is better	critical > 20
PGHF14-012	Long transactions blocking WAL recycle	integer (seconds)	lower is better	critical > 300
PGHF14-013	WAL accumulation since checkpoint	numeric (bytes)	lower is better	critical > 2147483648 · warning > 1073741824

C15 — Replication Health

Check	Name	Result	Judged as	Default threshold
PGHF15-001	WAL receiver last-message age	numeric (seconds)	lower is better	warning > 300
PGHF15-002	Inactive physical replication slots	integer (count)	lower is better	critical > 0
PGHF15-003	Standby query conflicts by type	jsonb	—	none yet — jsonb result needs a custom evaluator