System settings

Manage instance configuration under System Settings. Changes apply immediately and are included in database backups; no database or .env editing is needed. Source filtering and execution policies belong to Trackers and Executors.

Global configuration

Setting	Default / range	Effect
BASE URL	Empty; non-empty values must be canonical absolute HTTPS URLs	Notification links, frontend base path, OIDC callbacks and repository webhook endpoints; include any sub-path
Time zone	UTC; valid IANA zone such as Asia/Shanghai	Date display, maintenance windows and times in notification messages
Log level	INFO; DEBUG/INFO/WARNING/ERROR	Backend log verbosity
OCI registry redirects	Disabled	Enable only when a registry serves content through redirects; redirects still pass security checks
Readiness defaults	See Health checks	Used when an executor does not override them
Release channel history	20; 1–1000	Version records kept per release channel
Executor runtime snapshots	10; 1–1000	Snapshots kept per executor; locked or claimed snapshots are protected, so the actual count may be higher
Backup retention	7; 1–100	Full backups kept in the data directory's backups folder
Automatic backup interval	0 (off); 0–8760 hours	Interval for automatic full backups, see Online backup

Clean now trims release history or snapshots to the current limit; release history is also pruned daily at 02:00 or after a maintenance window ends. Check which records you need before lowering limits; pruning is not a backup. See Reverse proxy and sub-paths for proxies and Backup, upgrade and keys for key rotation.

Runtime operation policy (API)

Configure config.operation_policy through the runtime connection API; there is no form yet. This is a config fragment; keep other keys when editing:

{
  "operation_policy": {
    "read_timeout_seconds": 20,
    "write_timeout_seconds": 90,
    "read_retries": 1
  }
}

Field	Default	Valid range (integer)
read_timeout_seconds	20 s	1–600
write_timeout_seconds	90 s	1–600
read_retries	1 additional attempt	0–3; 0 disables retries

Call path	Policy application
Portainer reads	Read timeout; bounded retries for transient transport failures
Portainer stack writes	Write timeout; no automatic replay
Docker / Podman SDK	Client network timeout uses the write timeout
Compose / Helm commands	Subprocesses use the write budget; no automatic replay
Native Kubernetes workload API	Not yet integrated

This is not a whole-update deadline. A timed-out write may already have executed on the server; see Update timeout.