Database
Managed databases
Managed database instances, their replicas, backups and recovery runs.
Tables on this page: backup · managed · recovery · replica. Generated from migration 0004_schema_comments; see How to read this.
backup
One completed managed-engine backup artifact recorded from the daemon's managed.backup result; unique per (managed_id, backup_id), cascades with the cluster.
| Column | Type | Null | Default | Description |
|---|---|---|---|---|
id (PK) | uuid | no | uuidv7() | Primary key (uuidv7(), time-ordered). |
created_at | timestamp(3) with time zone | no | now() | Row creation time. |
managed_id | uuid | no | FK → managed.id (on delete cascade). | |
backup_id | text | no | Daemon-minted bk_ plus hex token that is also the artifact filename on the host; unique per managed engine, not globally. | |
size_bytes | bigint | no | Artifact size in bytes as reported by the daemon after writing the dump; re-checked before a restore. | |
checksum | text | no | Lowercase SHA-256 hex digest of the artifact computed by the daemon; a restore refuses on mismatch. | |
database | text | yes | Database name for a single-database backup; null for an instance-scope backup. | |
path | text | no | Absolute artifact path on the primary server's filesystem as reported by the daemon. |
Constraints and indexes
- Index
idx_backup_managed_id_created_at: (managed_id,created_at) - Unique index
uniq_backup_managed_backup_id: (managed_id,backup_id) - Check
backup_id_format_check:backup_id ~ '^[A-Za-z0-9_-]+$' - Check
backup_checksum_format_check:checksum ~ '^[a-f0-9]{64}$' - Check
backup_size_bytes_check:size_bytes >= 0
managed
One managed database cluster per environment (1:1 via unique environment_id); created by the managed create route, status projected by the command consumer.
| Column | Type | Null | Default | Description |
|---|---|---|---|---|
id (PK) | uuid | no | uuidv7() | Primary key (uuidv7(), time-ordered). |
created_at | timestamp(3) with time zone | no | now() | Row creation time. |
updated_at | timestamp(3) with time zone | no | now() | Last write time; the ORM sets it to now() on every update. |
metadata | jsonb | yes | Residual cluster facts: rootPrincipalId and rootUsername set at create, plus host, port and error written by the command consumer from daemon results. | |
options | jsonb | yes | Operator settings jsonb with keys settings (engine ManagedSettings) and databases (string array), written by the managed routes and parsed per engine spec. | |
environment_id | uuid | no | FK → environment.id (on delete cascade). Owning environment; exactly one managed row per environment (unique index), cascade-deleted with it. | |
server_id | uuid | yes | FK → server.id (on delete restrict). Placement pin of the primary member's host, copied from the environment at create and re-set by the consumer only on a primary-member apply success. | |
name | text | yes | Operator label for the cluster (1-255 chars of letters, digits, space, dot, underscore, hyphen); null when unnamed. | |
engine | text | no | Catalog engine code set once at create from ManagedEngineSpec: postgres, mysql, mariadb, redis or clickhouse (only the first three are creatable). | |
status | text | yes | Lifecycle state: provisioning at create, applying while a command is queued, then daemon-observed ready, stopped or failed projected by the consumer. |
Constraints and indexes
- Index
idx_managed_environment_id: (environment_id) - Index
idx_managed_server_id: (server_id) - Index
idx_managed_engine: (engine) - Unique index
managed_environment_id_unique: (environment_id) - Check
managed_name_format_check:("managed"."name" IS NULL) OR (((char_length(("managed"."name")::text) >= 1) AND (char_length(("managed"."name")::text) <= 255)) AND (("managed"."name")::text ~ '^[A-Za-z0-9 ._-]+$'::text)) - Check
managed_status_check:status IS NULL OR status IN ('provisioning','applying','ready','stopped','failed')
recovery
Durable HA journal entry for one managed cluster recovery (failover, switchover or DR); at most one non-terminal row per managed_id, written by the HA flow.
| Column | Type | Null | Default | Description |
|---|---|---|---|---|
id (PK) | uuid | no | uuidv7() | Primary key (uuidv7(), time-ordered). |
created_at | timestamp(3) with time zone | no | now() | Row creation time. |
updated_at | timestamp(3) with time zone | no | now() | Last write time; the ORM sets it to now() on every update. |
metadata | jsonb | yes | Fencing and progress facts: fencingEpoch, fenceCommandIds, promoteCommandId, haPresent, fenced, drainApplied, lagBytes, blockedReason. | |
options | jsonb | yes | Reserved jsonb; no current code path writes it. | |
managed_id | uuid | no | FK → managed.id (on delete cascade). | |
kind | text | no | automatic-failover, switchover (planned promotion of a failover replica) or disaster-recovery (promotion of a remote read replica). | |
source_primary_member_id | uuid | no | replica.id of the primary being replaced, stored without an FK so deleting the member cannot block the journal. | |
target_member_id | uuid | yes | replica.id of the member being promoted, stored without an FK; null when an automatic failover was blocked with no eligible candidate. | |
state | text | no | Journal phase: detecting, fencing, promoting, repointing, reconciling-ingress, verifying, then terminal completed, failed or blocked. | |
started_at | timestamp(3) with time zone | no | now() | When the recovery was opened; set explicitly by insertRecovery together with the initial state. |
completed_at | timestamp(3) with time zone | yes | When the recovery reached a terminal state; null while still in flight. |
Constraints and indexes
- Index
idx_recovery_managed_id: (managed_id) - Unique index
uniq_recovery_inflight_managed: (managed_id) where"recovery"."state" NOT IN ('completed','failed','blocked') - Check
recovery_kind_check:"recovery"."kind" IN ('automatic-failover','switchover','disaster-recovery') - Check
recovery_state_check:"recovery"."state" IN ('detecting','fencing','promoting','repointing','reconciling-ingress','verifying','completed','failed','blocked')
replica
One server's membership in a managed cluster (primary plus replicas); exactly one primary per managed_id, written by the member lifecycle and the consumer.
| Column | Type | Null | Default | Description |
|---|---|---|---|---|
id (PK) | uuid | no | uuidv7() | Primary key (uuidv7(), time-ordered). |
created_at | timestamp(3) with time zone | no | now() | Row creation time. |
updated_at | timestamp(3) with time zone | no | now() | Last write time; the ORM sets it to now() on every update. |
metadata | jsonb | yes | Daemon-observed replication health (state, observedAt, optional lagBytes, lagSeconds) written by the consumer after apply and lifecycle commands. | |
options | jsonb | yes | Reserved jsonb; no current code path writes it. | |
managed_id | uuid | no | FK → managed.id (on delete cascade). | |
server_id | uuid | no | FK → server.id (on delete restrict). Host running this member; one member per server per cluster and a private port is unique per server; deleting the server is restricted. | |
role | text | no | 'primary' | primary or replica; a partial unique index allows one primary per cluster and the consumer swaps roles on promote and failover. |
replica_class | text | yes | failover (same datacenter, promotable) or read (any org server, never auto-promoted); null on the primary and ignored when role is primary. | |
is_read_eligible | boolean | no | false | Whether the listener may route read-only logins to this member; true on the primary, operator-chosen on replicas (API field readEligible). |
ordinal | integer | no | 1 | 1-based member ordinal, unique per cluster; the primary is 1 and it mirrors the engine service's container ordinal. |
replication_transport | text | yes | Resolved private path from this member to the primary: local, datacenter, fabric or public; null until resolved or on the primary. | |
private_port | integer | yes | Host port 45000-45999 published on the member's private address for replication and ProxySQL backends; allocated per server, null for single-member clusters. | |
status | text | yes | Per-member state: provisioning, applying, ready, stopped, failed or needs_resync (set on demote or a failed fence, never auto-cleared). |
Constraints and indexes
- Unique
uniq_replica_managed_ordinal: (managed_id,ordinal) - Unique
uniq_replica_managed_server: (managed_id,server_id) - Index
idx_replica_managed_id: (managed_id) - Index
idx_replica_server_id: (server_id) - Unique index
uniq_replica_primary: (managed_id) where"replica"."role" = 'primary' - Unique index
uniq_replica_server_private_port: (server_id,private_port) where"replica"."private_port" IS NOT NULL - Check
replica_role_check:"replica"."role" IN ('primary','replica') - Check
replica_replica_class_check:"replica"."replica_class" IS NULL OR "replica"."replica_class" IN ('failover','read') - Check
replica_ordinal_positive_check:"replica"."ordinal" >= 1 - Check
replica_transport_check:"replica"."replication_transport" IS NULL OR "replica"."replication_transport" IN ('local','fabric','datacenter','public') - Check
replica_status_check:status IS NULL OR status IN ('provisioning','applying','ready','stopped','failed','needs_resync')
Edit on GitHub
Last updated on