Kronika records Linux metrics, PostgreSQL statistics, query plans and events from PostgreSQL/PgBouncer logs. The collector runs on the monitored Linux machine or records only PostgreSQL data from a local or remote server. The web interface shows resource use, processes, queries and locks during a selected hour, and how they changed over time.
A recorded hour, 21 September 2026, 08:59:46–09:59:46 UTC: Processes · Statements · Plans · Host · Events.
Install Kronika 1.2.4 from a Linux release archive, or build from source.
Choose local to record Linux and, optionally, PostgreSQL in the same VM or pod.
Choose postgresql for a remote server or when you only need database metrics.
Start collecting Linux metrics:
sudo /usr/local/bin/kronika-collector \
--storage-dir /var/lib/kronika
For PostgreSQL on the collector machine, supply its connection string with
--pg-dsn when starting collector. Use a PostgreSQL account with the
monitoring privileges.
sudo /usr/local/bin/kronika-collector \
--storage-dir /var/lib/kronika \
--pg-dsn 'host=127.0.0.1 port=5432 user=kronika_monitor password=replace-with-password dbname=postgres sslmode=disable'In this default local mode, the connection also discovers the current
PostgreSQL log file. If that file is readable on the collector host, --pg-log
is not needed.
Leave --postgres-effective-cpus and KRONIKA_POSTGRES_EFFECTIVE_CPUS unset
for this local machine: CPU capacity comes from its recorded CPU snapshots.
sudo install -d -m 0700 -o "$(id -u)" /var/lib/kronika
/usr/local/bin/kronika-collector \
--mode postgresql \
--storage-dir /var/lib/kronika \
--pg-dsn 'host=pg.example.net port=5432 user=kronika_monitor password=replace-with-password dbname=postgres'If the PostgreSQL CPU capacity is known, add --postgres-effective-cpus 4,
replacing 4 with its CPU count.
Without it, SQL metrics remain available. PostgreSQL Health is unknown.
See collector configuration.
To collect from several PostgreSQL servers, run a kronika-collector process
for each server with its DSN and a separate storage directory. See the
two-server example.
Run kronika-web with the collector’s data directory.
sudo /usr/local/bin/kronika-web --storage-dir /var/lib/kronika \
--listen 0.0.0.0:8080/usr/local/bin/kronika-web --storage-dir /var/lib/kronika \
--listen 0.0.0.0:8080Open http://<server-ip>:8080.
To require sign-in, add --user kronika --password 'replace-with-a-random-password'.
Options override their environment fallbacks; existing KRONIKA_WEB_* service
settings continue to work. See kronika-web --help for the full option list.
Systemd setup covers running both programs as services and changing an existing service's configuration.
For a PostgreSQL workload with roughly 500 tables and 3,000 indexes, estimate about 200 MB of compressed recordings per day. Volume depends on collection intervals and the number of recorded objects and distinct queries.
The default storage limit is 2 GiB (--retention), including journals and
indexes. When the limit is exceeded, the collector removes the oldest finished
recordings and their indexes.
auto and auto:P instead set a used-space percentage target for the whole
backing filesystem. See storage configuration
for the rotation rules and automatic mode.
| Domain | What you can inspect | Reference |
|---|---|---|
| Processes | Command, state and process number (PID), CPU use, memory and disk reads/writes, process tree and hourly activity. | Linux metrics |
| Host | CPU, memory, time waiting for resources (PSI), network and disks, free space and device relationships, cgroup resource limits and use. | Linux metrics |
| PostgreSQL sessions | Overview, Activity, Locks, Vacuum: session states and waits, blocking chains, query/transaction durations and table cleanup progress. | PostgreSQL metrics |
| Queries and plans | Statements and Plans: calls, execution/planning time, page and temporary-file reads, write-ahead log (WAL) output, SQL and plan text. | PostgreSQL metrics |
| PostgreSQL objects | Databases, Tables and Indexes: settings, size, reads and changes, maintenance and transaction ages. Grouping by database, schema and tablespace. | PostgreSQL metrics |
| Events | Grouped PostgreSQL/PgBouncer log events, occurrences, durations and recorded context, metric marks. | Views and controls |
| Time and charts | Choose an hour and a time within it. View changes, activity maps, totals and the distribution of measurements. | Time and calculations |
Views and controls explains how to select measurements, group, search and sort rows, inspect details in Inspector, view charts and export. The operator guide contains four worked examples from the 5 September recording.
The default collection intervals are 5 seconds for processes, 10 seconds for core Linux metrics, 30 seconds for PostgreSQL metrics, and 300 seconds for tables and indexes. Collector configuration defines source selection, intervals, access permissions and removal of old recordings.
The web server serves the browser, HTTP API and MCP at one address and port. The AI panel provides connection settings. MCP tools return values at a chosen time, objects ranked by a measurement, field descriptions, events and row details.
Export saves a selected interval of your recording as one interactive HTML file. It includes the interface and data, so tables, search and charts work without a server or network connection.
kronika-dump inspects storage and extracts an interval into a standalone ZMS recording, and kronika-report converts that recording into HTML.
- Setup: Install · Archives and CI · Services · Storage failures · Source build
- Reference: Controls · Time · Linux · PostgreSQL · MCP
- Programs: Collector · Web · Dump · Report
- Recorded fields: Linux · PostgreSQL metrics · PostgreSQL events · PgBouncer events
- Development: Libraries · Segment format · Development demo