Retention, backups, and capacity

Plan storage from two directions: PostgreSQL metadata recovery and ClickHouse telemetry retention. Valkey holds runtime state and can be rebuilt.

Inputs

• Storage capacity targets
• Recovery point objectives for PostgreSQL and ClickHouse
• Project-level retention decisions

Retention and backups

Project retention is stored in metadata and controls telemetry TTL columns:

Setting	Default
Spans retention	30 days
Logs retention	14 days
Metrics retention	30 days
Histograms retention	14 days
Exemplars retention	30 days

Environment growth is controlled with:

Setting	Default
auto_register_environments	true
max_environments	32
ephemeral_slug_patterns	empty

Back up PostgreSQL metadata for users, sessions, workspaces, projects, environments, roles, tokens, dashboards, saved queries, channels, alert rules, silences, feature flag definitions, flag targeting, segments, flag change history, and audit logs.

Feature flag evaluation telemetry is ordinary telemetry once emitted. It follows the project’s metric retention, while flag change logs follow the project’s log retention.

Back up ClickHouse if telemetry recovery matters for your operations. Otherwise, ClickHouse can be treated as time-bounded telemetry storage.

Check capacity

Run:

oriel --config /etc/oriel/oriel.toml doctor

Review disk headroom warnings and database-specific monitoring.

Recover or adjust

• Restore PostgreSQL before starting Oriel when control-plane state is lost.
• Restore ClickHouse when historical telemetry must be preserved.
• If environment caps reject new preview environments, prune stale environments or raise max_environments.
• If metric series caps reject new series, inspect recent instrumentation cardinality before raising limits.