Archive Metrics Diagnostics endpoint, Archive Metrics run endpoint and fixes to the internal run method

obp-api/src/main/scala/bootstrap/liftweb/Boot.scala

@@ -103,7 +103,7 @@ import code.metadata.tags.MappedTag
 import code.metadata.transactionimages.MappedTransactionImage
 import code.metadata.wheretags.MappedWhereTag
 import code.methodrouting.MethodRouting
-import code.metrics.{ConnectorTrace, MappedConnectorMetric, MappedMetric, MetricArchive}
+import code.metrics.{ConnectorTrace, MappedConnectorMetric, MappedMetric, MetricArchive, MetricsArchiveRun}
 import code.migration.MigrationScriptLog
 import code.model._
 import code.model.dataAccess._
@@ -994,6 +994,7 @@ object ToSchemify extends MdcLoggable {
     TransactionRequestAttribute,
     MappedMetric,
     MetricArchive,
+    MetricsArchiveRun,
     MapperAccountHolders,
     MappedEntitlement,
     MappedConnectorMetric,

obp-api/src/main/scala/code/api/util/ApiRole.scala

@@ -448,6 +448,12 @@ object ApiRole extends MdcLoggable{
   case class CanGetDatabasePoolInfo(requiresBankId: Boolean = false) extends ApiRole
   lazy val canGetDatabasePoolInfo = CanGetDatabasePoolInfo()
 
+  case class CanGetMetricsDiagnostics(requiresBankId: Boolean = false) extends ApiRole
+  lazy val canGetMetricsDiagnostics = CanGetMetricsDiagnostics()
+
+  case class CanCreateMetricsArchiveRun(requiresBankId: Boolean = false) extends ApiRole
+  lazy val canCreateMetricsArchiveRun = CanCreateMetricsArchiveRun()
+
   case class CanGetConnectorHealth(requiresBankId: Boolean = false) extends ApiRole
   lazy val canGetConnectorHealth = CanGetConnectorHealth()
 

obp-api/src/main/scala/code/api/v7_0_0/Http4s700.scala

@@ -7,7 +7,7 @@ import code.api.Constant._
 import code.api.ResourceDocs1_4_0.SwaggerDefinitionsJSON._
 import code.api.util.APIUtil.{EmptyBody, _}
 import code.api.util.{APIUtil, ApiRole, CallContext, CustomJsonFormats, Glossary, NewStyle}
-import code.api.util.ApiRole.{canCreateEntitlementAtAnyBank, canCreateEntitlementAtOneBank, canCreateOrganisation, canCreateRoutingScheme, canCreateTestEmail, canDeleteEntitlementAtAnyBank, canDeleteOrganisation, canDeleteRoutingScheme, canGetAccountAccessTrace, canGetAnyOrganisation, canGetAnyUser, canGetCacheConfig, canGetCacheInfo, canGetCacheNamespaces, canGetConnectorHealth, canGetCustomersAtOneBank, canGetDatabasePoolInfo, canGetMigrations, canUpdateBankSupportedRoutingScheme, canUpdateOrganisation, canUpdateRoutingScheme, canUpdateSystemView}
+import code.api.util.ApiRole.{canCreateEntitlementAtAnyBank, canCreateEntitlementAtOneBank, canCreateMetricsArchiveRun, canCreateOrganisation, canCreateRoutingScheme, canCreateTestEmail, canDeleteEntitlementAtAnyBank, canDeleteOrganisation, canDeleteRoutingScheme, canGetAccountAccessTrace, canGetAnyOrganisation, canGetAnyUser, canGetCacheConfig, canGetCacheInfo, canGetCacheNamespaces, canGetConnectorHealth, canGetCustomersAtOneBank, canGetDatabasePoolInfo, canGetMetricsDiagnostics, canGetMigrations, canUpdateBankSupportedRoutingScheme, canUpdateOrganisation, canUpdateRoutingScheme, canUpdateSystemView}
 import code.api.util.CommonsEmailWrapper
 import code.model.dataAccess.AuthUser
 import code.api.util.ApiTag._
@@ -3240,6 +3240,126 @@ object Http4s700 {
       http4sPartialFunction = Some(factoryResetSystemView)
     )
 
+    // Route: GET /obp/v7.0.0/management/system/diagnostics/metrics
+    //
+    // Operator diagnostic for the metrics-archiving pipeline. Reports the
+    // archiving props plus row counts and the oldest/newest record in both the
+    // `metric` and `metricarchive` tables, then runs integrity checks that
+    // surface whether MetricsArchiveScheduler is keeping each table inside its
+    // configured retention window. Intended for use from the API Manager.
+    val getMetricsDiagnostics: HttpRoutes[IO] = HttpRoutes.of[IO] {
+      case req @ GET -> `prefixPath` / "management" / "system" / "diagnostics" / "metrics" =>
+        EndpointHelpers.withUser(req) { (_, _) =>
+          Future {
+            JSONFactory700.createMetricsAndArchiveMetricsDiagnosticsJsonV700()
+          }
+        }
+    }
+
+    resourceDocs += ResourceDoc(
+      implementedInApiVersion,
+      nameOf(getMetricsDiagnostics),
+      "GET",
+      "/management/system/diagnostics/metrics",
+      "Get Metrics and Archive Metrics Diagnostics",
+      s"""Diagnostic for the metrics-archiving pipeline (`MetricsArchiveScheduler`).
+         |
+         |Returns:
+         |
+         |* `config` — the relevant props and their *effective* values after the
+         |  scheduler's floors are applied (`retain_metrics_days` floored to 60,
+         |  `retain_archive_metrics_days` floored to 365): `write_metrics`,
+         |  `enable_metrics_scheduler`, `retain_metrics_scheduler_interval_in_seconds`,
+         |  `retain_metrics_days`, `retain_archive_metrics_days`,
+         |  `retain_metrics_move_limit`.
+         |* `metric` — row count and oldest/newest record (date + age in days) of
+         |  the live `metric` table.
+         |* `metric_archive` — the same for the `metricarchive` table.
+         |* `last_run` / `last_successful_run` — the most recent (and most recent
+         |  successful) scheduler run from the `metricsarchiverun` audit log, including
+         |  rows moved, rows deleted, duration and success. Absent if no run has been
+         |  recorded yet.
+         |* `checks` — a list of integrity checks, each with a `status` of `OK`,
+         |  `WARNING`, or `ERROR`:
+         |    * `write_metrics_enabled`
+         |    * `metrics_scheduler_enabled`
+         |    * `metric_oldest_within_retention` — flags if the oldest live metric
+         |      is older than the retention window (move job not keeping up / stopped).
+         |    * `archive_oldest_within_retention` — flags if the oldest archived
+         |      metric is older than the archive retention (cleanup not keeping up / stopped).
+         |    * `archive_recently_updated` — flags if a backlog exists but the newest
+         |      archived record is stale (move job stopped).
+         |    * `archive_job_last_run` — reports the outcome of the most recent run
+         |      from the run log (errors if the last run failed; warns if none recorded).
+         |* `everything_as_expected` — `true` only when every check is `OK`.
+         |
+         |${userAuthenticationMessage(true)}""".stripMargin,
+      EmptyBody,
+      JSONFactory700.metricsAndArchiveMetricsDiagnosticsJsonV700Example,
+      List($AuthenticatedUserIsRequired, UserHasMissingRoles, UnknownError),
+      apiTagMetric :: apiTagSystem :: apiTagApi :: Nil,
+      Some(List(canGetMetricsDiagnostics)),
+      http4sPartialFunction = Some(getMetricsDiagnostics)
+    )
+
+    // Route: POST /obp/v7.0.0/management/system/diagnostics/metrics/run
+    //
+    // Manually trigger one metrics-archive run. This calls the exact same
+    // `MetricsArchiveScheduler.runOnce()` the timer uses, so it honours the same
+    // concurrency lock (won't start if a run is already in progress), the same
+    // retention props/floors, and records the run in the `metricsarchiverun` log.
+    // The run executes synchronously and may take a while for large backlogs
+    // (it moves up to `retain_metrics_move_limit` rows).
+    val triggerMetricsArchiveRun: HttpRoutes[IO] = HttpRoutes.of[IO] {
+      case req @ POST -> `prefixPath` / "management" / "system" / "diagnostics" / "metrics" / "run" =>
+        EndpointHelpers.withUser(req) { (user, _) =>
+          Future {
+            val outcome = code.scheduler.MetricsArchiveScheduler.runOnce()
+            logger.info(s"AUDIT triggerMetricsArchiveRun: user_id=${user.userId} outcome=${outcome.getClass.getSimpleName}")
+            JSONFactory700.createTriggerMetricsArchiveRunResponseJsonV700(outcome)
+          }
+        }
+    }
+
+    resourceDocs += ResourceDoc(
+      implementedInApiVersion,
+      nameOf(triggerMetricsArchiveRun),
+      "POST",
+      "/management/system/diagnostics/metrics/run",
+      "Trigger a Metrics Archive Run",
+      s"""Manually run the metrics-archiving job once, on demand.
+         |
+         |This invokes the **same** code path as the scheduled
+         |`MetricsArchiveScheduler` run, so it respects all the same checks:
+         |
+         |* **Concurrency lock** — if an archive run is already in progress (the
+         |  `JobScheduler` lock is held, on this or another node), no new run is
+         |  started and the response `status` is `skipped_already_in_progress`.
+         |* **Retention rules** — moves `metric` rows older than the effective
+         |  `retain_metrics_days` (floored to 60) to `metricarchive`, up to
+         |  `retain_metrics_move_limit` rows; then deletes `metricarchive` rows
+         |  older than the effective `retain_archive_metrics_days` (floored to 365).
+         |* **Run log** — the outcome is written to the `metricsarchiverun` audit
+         |  log, exactly as a scheduled run would be.
+         |
+         |Response fields:
+         |* `status` — `completed` (a run executed — inspect `run.success`) or
+         |  `skipped_already_in_progress`.
+         |* `message` — human-readable summary.
+         |* `run` — the recorded run (run id, counts, duration, success, remark);
+         |  absent when skipped.
+         |
+         |Note: the run executes synchronously, so a large backlog may take a while.
+         |
+         |${userAuthenticationMessage(true)}""".stripMargin,
+      EmptyBody,
+      JSONFactory700.triggerMetricsArchiveRunResponseJsonV700Example,
+      List($AuthenticatedUserIsRequired, UserHasMissingRoles, UnknownError),
+      apiTagMetric :: apiTagSystem :: apiTagApi :: Nil,
+      Some(List(canCreateMetricsArchiveRun)),
+      http4sPartialFunction = Some(triggerMetricsArchiveRun)
+    )
+
     // Enabled only in Lift test mode (Props.testMode == true, i.e. -Drun.mode=test).
     // Props.testMode is set from the JVM system property before any props file loads,
     // so it is reliably available at object-initialization time unlike file-based props.