PostgreSQL Health Check

Use Health Check to see what is wrong with one PostgreSQL instance now, what to do about it, and what fills up next.

Before you begin

  • You need the View PostgreSQL and View HealthCheck permissions.
  • Scops checks the instance every 30 minutes. The page reads the newest result every minute.
  • The disks and the server log need an OS credential on the instance (see PostgreSQL permissions). Without one, these checks show not checked. With Manage databases, the Storage tab gives the link Add an OS credential.
  • To show one database, select it in the database bar. The page then shows the findings of that database and of the instance.

Run the check now

  1. At the top of the page, click the refresh button next to the time of the last check.

Scops reads every area again and shows the result when it ends. If the instance does not answer, the page keeps the last result and tells you why.

Read the cards

The cards at the top are the tabs: Overview, Vacuum, Storage, Replication, Indexes and IDs and Configuration. Each card shows the worst state of its area, one figure and one line. Click a card to open its tab.

CardFigure
OverviewThe count of problems, and a bar of all checks by state.
VacuumThe oldest transaction ID age, as a percent of the limit.
StorageThe used percent of the fullest file system.
ReplicationThe lag of the standby that is furthest behind. On a standby, its own delay.
Indexes and IDsThe used percent of the ID column nearest its limit.
ConfigurationThe count of problems, and the connections in use against max_connections.

Work through the checklist

The Overview tab lists every check in three groups:

  • Needs attention: the critical and warning checks, the most urgent first.
  • Could be better: advice that is not urgent, for example space to win back.
  • Healthy: the checks that passed, by area, and the checks that Scops could not read.
  1. Click a row. It shows Why it matters and What to do, and often a command and the first or largest items.

  2. To see more, click Open its history or Open the sessions in Live Activity, when the row has it.

  3. To see all rows of the area, click the link at the end, for example Open Vacuum: all databases.

Read the area tabs

Each area tab shows its tables, the rows with a problem first. Problem says what is wrong, and Since says from when. Click a row with a problem to read why it matters and what to do.

TabTables
VacuumAll databases: the transaction ID age of each database against the limit, its dead rows, its last autovacuum and its size.
StorageDisks: each file system, how full it is and what it holds. All databases: the size, tablespace and 30-day growth of each database. What PostgreSQL keeps: the data, the WAL, the server log, the temporary files and each tablespace, with their place and size.
ReplicationStandbys, Lag history, Replication slots and WAL retained by slots.
Indexes and IDsID columns nearest their limit: the sequences and their columns. Indexes never used, the largest first. Duplicate indexes.
ConfigurationSettings: the settings that matter, with their values. Server log: the severe lines of the last 24 hours, by kind.

To find a row in a long table, type in its search box, for example Database name. Indexes never used shows the first 20 rows. Click Show all to see the rest.

See when something fills up

Four tables show when a value reaches its limit at the pace of its history:

  • Full in, in Disks and All databases on the Storage tab.
  • Reaches limit, in All databases on the Vacuum tab and in ID columns nearest their limit.
The cell saysMeaning
For example full in ~9 d or in ~9 dThe days that are left at the pace of the last days. The cell is amber or red when the date is near.
full now, at the limitNo room is left.
not growing, not risingNo growth in the history.
vacuum firstThe age is under autovacuum_freeze_max_age. Autovacuum freezes the oldest tables when they reach it, long before the limit.
no historyScops does not have enough history yet.
> 1 yearMore than one year is left.

When Radar Engine has an open finding with a date for the row, the cell shows the date of Radar. The line above the table then shows risks as of and a time. To see the values over the last days, the pace and the limit, click the row.

Follow the transaction ID age and the sequences

  1. On the Vacuum or the Indexes and IDs tab, click History.

  2. Select the tab: Transaction ID Age or Sequences.

  3. Select a period: 1D, 1W or 1M.

The chart shows the oldest databases or the fullest sequences. The table gives the value now and its change in the period. A line that only climbs is a database that vacuum does not freeze, or a sequence that will fail inserts.

Check replication

The Replication tab has these tables:

TableWhat it shows
StandbysEach standby with its state, its mode (for example synchronous or asynchronous), its lag and how much WAL it is behind.
Lag historyThe lag of each standby over the period, at its peak in each interval. On a standby, the average of its own delay. A line tells you when a standby was 30 seconds or more behind.
Replication slotsEach slot with its kind, its database and the WAL that it holds. Read shows active, or how long no consumer has read the slot.
WAL retained by slotsThe WAL that the slots hold over the period, against max_wal_size.

To change the period of the two charts, select Last 24 hours, Last 7 days (the default) or Last 31 days. To see the charts with the other replication charts, click Open the Metrics page.

Read the server log

The Server log table on the Configuration tab counts the severe lines of the last 24 hours, by kind:

  • Damaged data
  • Crash
  • No space left on the device
  • Out of memory
  • Connections refused

Newest shows the time of the last line of each kind.

Scops reads the log over SSH. It looks for the English messages of PostgreSQL only. If lc_messages is a different language, the table tells you to set lc_messages = 'C'. To read the log lines, open Log Watch from the row menu of the instance on the Databases page. You need Manage databases.

Updated November 22, 2026