api/me.ahoo.wow.api.ledger/Ledger

Ledger

The failure ledger across services: each service keeps its own ledger, atomic with its checkpoints, and publishes its changes to the ops context, which aggregates them; an operator's action goes back to the owning service.

Publishing: a service sends each change of an entry as a command to the aggregate PARKED of the ops context (OPS_CONTEXT by default), through that context's command topic: RECORD_PARKED with the LedgerEntry as its member ENTRY while the entry is parked, RECORD_RESOLVED with the LedgerResolution as its member RESOLUTION once it is not. The aggregate's id is parkedId: one per service, subscription and parked aggregate.
Acting: the ops context sends the system commands RETRY, SKIP, MARK_RECOVERABLE and APPLY_RETRY_SPEC (RetryParked, SkipParked, MarkParkedRecoverable, ApplyParkedRetrySpec) to the command topic of the context that holds the entry (LedgerEntry.context), addressed to targetOf that context and the entry's ops id. The owning service authorizes each as the system command of that name on the parked aggregate, as it does the same action on its own operator routes, and runs it.

Their names sit in the system namespace NAMESPACE, as the lifecycle commands' do.

Properties

APPLY_RETRY_SPEC

const val APPLY_RETRY_SPEC: String

COMMANDS

@JvmField

val COMMANDS: Map<String, Class<out LedgerCommand>>

The ledger's system commands, by their names.

ENTRY

const val ENTRY: String

The member of RECORD_PARKED's body that holds its LedgerEntry.

MARK_RECOVERABLE

const val MARK_RECOVERABLE: String

NAMESPACE

const val NAMESPACE: String

The namespace of the ledger's system commands.

OPS_CONTEXT

const val OPS_CONTEXT: String

The context a service publishes its ledger to unless it is configured otherwise.

PARKED

const val PARKED: String

The aggregate of the ops context that records one parked entry of one service.

RECORD_PARKED

const val RECORD_PARKED: String

The simple names of the commands a service sends to PARKED: their full names are <ops>.parked.<name>.

RECORD_RESOLVED

const val RECORD_RESOLVED: String

RESOLUTION

const val RESOLUTION: String

The member of RECORD_RESOLVED's body that holds its LedgerResolution.

RETRY

const val RETRY: String

SKIP

const val SKIP: String

Functions

nameOf

@JvmStatic

fun nameOf(type: Class<*>): String?

The system name of type, a ledger command, or null when it is none.

parkedId

@JvmStatic

fun parkedId(service: String, context: String, subscription: String, ref: AggregateRef): String

The id of the ops aggregate recording the entry of ref that subscription, of context, parked in service: <service>:<context>:<subscription>:<aggregate context>.<aggregate>:<tenant>:<id>, so one service's entry of one aggregate is one ops aggregate, whichever node publishes it, and two contexts' subscriptions of one name keep entries of their own. Each field has its % and : percent-escaped (%25, %3A, Escapes), and the aggregate type's context its . too (Escapes.typeText), so two entries never share an id however their names join: tenant a:b with id c and tenant a with id b:c stay apart, as do context a.b's aggregate c and context a's b.c.

targetOf

@JvmStatic

fun targetOf(context: String, parkedId: String): AggregateRef

Where a system command of the ledger on the entry parkedId names (Ledger.parkedId) is sent: the command topic of context, the context that holds the entry, keyed by the entry rather than by the parked aggregate, so a command on an entry of the context's own intake is not held behind the very aggregate it parked, and the commands of one entry keep their order.