RIM
DocsOperatorsBackground jobs

Background jobs

rim-server runs these jobs inside its own process on a timer; there is no cron to set up. Each starts with the server and stops when it shuts down. A run that fails logs an error and the next run tries again. More than one server process on the same database each runs every job; the purges are idempotent, so a second run finds nothing left to do.

Job Interval Setting What it does
Presence and typing 30 seconds none Moves users without a heartbeat to away after 90 seconds and offline after 180, and drops stale typing indicators.
Pipeline evaluator on every event none Runs the workspace pipelines that match each event.
Notification digest 15 minutes RIM_NOTIFICATION_DIGEST_INTERVAL_SECS (at least 60) Emails due notification digests. Off when the email feature is disabled.
Notification and MLS retention hourly none Deletes notifications older than 90 days and the MLS rows the encryption retention rules let go.
Server-to-server upkeep every minute RIM_RELAY_RETENTION_DAYS (default 30) for the relay queue Deletes expired server-to-server nonces, relay queue items and DM requests past retention, outbox rows past their retention, old seen-request records and ended delegations; picks up a server key or franking key rotated by the CLI.
Revocation fan-out 5 seconds runs while server-to-server traffic is on Sends device revocations due for another attempt to the hosts holding the person.
Outbox writer and sender on every event runs while server-to-server traffic is on Turns events into signed envelopes and delivers each relay's outbox in order.
Metrics sampler 30 seconds runs when RIM_METRICS_ADDR is set Reads the relay queue, the database pool and pending migrations into gauges (docs/operations/metrics.md).
Audit retention daily, first a minute after boot RIM_AUDIT_RETENTION_DAYS; unset, the job does not run Exports and deletes command audit rows past retention (docs/operations/audit-retention.md).
Stale attachment cleanup daily, first 5 minutes after boot none Marks attachment uploads still pending 15 minutes after they started as failed, so they stop counting as in progress. Logs the count and adds it to rim_attachments_cleaned_total.
IP deny list 30 seconds, first loaded at boot none Reads the IP blocks in force, so a block or unblock made with rim_server ip applies within 30 seconds, and deletes expired blocks. A read that fails keeps the list it had.
Workspace state expiry hourly, first 5 minutes after boot none Deletes workspace state entries whose time to live ran out. Logs the count and adds it to rim_workspace_state_purged_total.
Conversation invitation expiry hourly, first 5 minutes after boot none Deletes pending invitations to encrypted conversations not accepted within 30 days. Logs the count and adds it to rim_conversation_invites_expired_total.
WebSocket session recheck RIM_WS_SESSION_RECHECK_SECS (default 30), first one interval after boot RIM_WS_SESSION_RECHECK_SECS; 0 turns it off Reads the sessions of this process's open WebSocket connections, one query per 500 sessions, and closes with 4401 the connections whose session is revoked, expired or gone: those ended in another process, from the command line or in the database. A read that fails is logged and closes nothing.
Account purge hourly, first 5 minutes after boot RIM_ACCOUNT_DELETION_GRACE_DAYS (default 14, 0 to 30) sets when a requested deletion is due Purges the accounts whose deletion grace period ended, each in its own transaction, at most 100 a run: cancels their subscriptions at other servers, deletes what they own and leaves a tombstone row. Writes an account.purged audit line with the counts per account; an account whose purge fails is logged and tried again next run. rim_server account purge [--dry-run] runs it once.