Operations

Scheduled jobs

The cron container runs supercronic against docker/supercronic/crontab. All commands are invoked as php /app/bin/console <name>.

Command	Schedule	Purpose
partitions:rotate	0 2 * * * — daily at 02:00 UTC	Create partitions up to 2 months ahead; drop partitions older than the retention window
rate_limits:cleanup	0 3 * * 0 — Sunday at 03:00 UTC	Delete expired rate-limit buckets from the database
bots:update	0 5 * * * — daily at 05:00 UTC	Refresh core.bot_ips from myip.ms bot IP list
bots:update-extra	30 5 * * * — daily at 05:30 UTC	Refresh core.bot_asns (brianhama bad-ASN list) and core.bot_cidrs (X4BNet datacenter + VPN CIDR lists)
db:vacuum-stats	30 3 * * * — daily at 03:30 UTC	VACUUM ANALYZE the current-month partition of each stats table
geoip:check	0 6 * * * — daily at 06:00 UTC	Check the GeoLite2 .mmdb files; logs a warning and exits non-zero if any is missing or older than 14 days (Telegram notification of stale GeoIP is delivered separately by the hourly telegram:alerts job)
stats:refresh	*/5 * * * * — every 5 minutes	Refresh the stats.clicks_hourly materialized view
telegram:digest	0 10 * * * — daily at 10:00 UTC	Send daily summary (clicks, conversions, revenue) to the configured Telegram chat
telegram:alerts	0 * * * * — every hour	Send system alerts (high bot rate, DB lag, etc.) to Telegram
db:backup	0 1 * * * — daily at 01:00 UTC	pg_dump in custom format; 14-day rotation
postback:deliver	* * * * * — every minute	Flush pending rows from core.postback_deliveries (outgoing S2S postbacks with exponential backoff retry)
inbox:flush	* * * * * — every minute	Drain stats.pixel_events_inbox → stats.pixel_events (GeoIP + UA enrichment); runs a 55-second inner loop so events surface within ~5–6 seconds
rrweb:flush	* * * * * — every minute	Drain stats.rrweb_inbox → gzipped stats.rrweb_chunks + upsert stats.rrweb_sessions; same 55-second inner loop so session replays surface within a few seconds

For details on how the bot detection lists are populated, see the Bot detection section in Traffic Filtering.

Partitions and retention

stats.clicks, stats.pixel_events, and stats.visitors_fingerprints are RANGE-partitioned by created_at in monthly intervals. Each partition also carries a BRIN index on created_at (except visitors_fingerprints, which uses a regular index on fp_hash). stats.rrweb_chunks (raw session replay data) is partitioned daily instead, since it is high-volume and short-lived.

partitions:rotate (runs daily at 02:00 UTC) does two things:

1. Creates ahead — ensures partitions exist for the current month and the next 2 months (3 total).
2. Drops expired — removes partitions whose data falls entirely outside the configured retention window.

Retention defaults (applied if no override is stored in core.settings):

Table	Default retention
stats.clicks	365 days
stats.pixel_events	90 days
stats.visitors_fingerprints	30 days
core.auth_events	180 days

To change retention windows, use the General tab at /admin/settings — see Settings & Notifications. Changes apply on the next partitions:rotate run.

To manually rotate partitions:

docker compose exec app php bin/console partitions:rotate

This is also required before running the integration test suite (make test-up calls it automatically against db-test).

Backups

db:backup (daily at 01:00 UTC) runs pg_dump --format=custom and writes the dump to /app/var/backups/ inside the app container. After each successful dump it scans for files older than 14 days and removes them, keeping at most 14 days of history.

To trigger a backup manually:

docker compose exec app php bin/console db:backup

To restore from a dump:

# List available dumps
docker compose exec app ls /app/var/backups/

# Restore (the 'yes' argument confirms the destructive operation)
docker compose exec app php bin/console db:restore <filename>.dump yes

No extra volume is needed for host-side copies: docker-compose.yml already bind-mounts ./var/backups on the host to /app/var/backups in both the app and cron containers, so dumps land in the repository directory as they are written.