CORDA-3195 Default behaviour of FlowHospital (WIP) (#2520)

* Add GP to flow hospital, and start working on a list of things the GP knows to be incurable.

* Only hospitalise SQL and Persistence Exceptions (let's see if that is enough?), also rename to DatabaseDentist.

* Disabled hospitalisation of SQL exceptions in flow retry tests

* Fix RPC exception handling test by not using PersistenceException

* Ignore flaky integration test

* Code review: Rename staff member and add testing annotation

* Revert compiler.xml

node/src/integration-test/kotlin/net/corda/node/flows/FlowRetryTest.kt

@@ -17,6 +17,7 @@ import net.corda.core.utilities.getOrThrow
 import net.corda.core.utilities.unwrap
 import net.corda.node.services.Permissions
 import net.corda.node.services.statemachine.FlowTimeoutException
+import net.corda.node.services.statemachine.StaffedFlowHospital
 import net.corda.testing.core.ALICE_NAME
 import net.corda.testing.core.BOB_NAME
 import net.corda.testing.core.singleIdentity
@@ -25,6 +26,7 @@ import net.corda.testing.driver.driver
 import net.corda.testing.node.User
 import org.assertj.core.api.Assertions.assertThatExceptionOfType
 import org.hibernate.exception.ConstraintViolationException
+import org.junit.After
 import org.junit.Before
 import org.junit.Test
 import java.lang.management.ManagementFactory
@@ -46,6 +48,12 @@ class FlowRetryTest {
         TransientConnectionFailureFlow.retryCount = -1
         WrappedTransientConnectionFailureFlow.retryCount = -1
         GeneralExternalFailureFlow.retryCount = -1
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.add { t -> true }
+    }
+
+    @After
+    fun cleanUp() {
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.clear()
     }
 
     @Test

node/src/integration-test/kotlin/net/corda/node/services/rpc/RpcExceptionHandlingTest.kt

@@ -140,7 +140,7 @@ class InitiatedFlow(private val initiatingSession: FlowSession) : FlowLogic<Unit
     @Suspendable
     override fun call() {
         initiatingSession.receive<String>().unwrap { it }
-        throw GenericJDBCException("Something went wrong!", SQLException("Oops!"))
+        throw ClientRelevantException("Something went wrong!", SQLException("Oops!"))
     }
 }
 

node/src/main/kotlin/net/corda/node/services/statemachine/StaffedFlowHospital.kt

@@ -7,10 +7,7 @@ import net.corda.core.flows.ReceiveTransactionFlow
 import net.corda.core.flows.StateMachineRunId
 import net.corda.core.flows.UnexpectedFlowEndException
 import net.corda.core.identity.Party
-import net.corda.core.internal.DeclaredField
-import net.corda.core.internal.ThreadBox
-import net.corda.core.internal.TimedFlow
-import net.corda.core.internal.bufferUntilSubscribed
+import net.corda.core.internal.*
 import net.corda.core.messaging.DataFeed
 import net.corda.core.utilities.contextLogger
 import net.corda.core.utilities.seconds
@@ -22,6 +19,7 @@ import java.sql.SQLTransientConnectionException
 import java.time.Duration
 import java.time.Instant
 import java.util.*
+import javax.persistence.PersistenceException
 import kotlin.math.pow
 
 /**
@@ -35,7 +33,8 @@ class StaffedFlowHospital(private val flowMessaging: FlowMessaging, private val
             DuplicateInsertSpecialist,
             DoctorTimeout,
             FinalityDoctor,
-                TransientConnectionCardiologist
+            TransientConnectionCardiologist,
+            DatabaseEndocrinologist
         )
     }
 
@@ -259,7 +258,6 @@ class StaffedFlowHospital(private val flowMessaging: FlowMessaging, private val
         NOT_MY_SPECIALTY
     }
 
-
     interface Staff {
         fun consult(flowFiber: FlowFiber, currentState: StateMachineState, newError: Throwable, history: FlowMedicalHistory): Diagnosis
     }
@@ -361,7 +359,6 @@ class StaffedFlowHospital(private val flowMessaging: FlowMessaging, private val
             return strippedStacktrace.isNotEmpty() &&
                     strippedStacktrace.first().className.startsWith(ReceiveTransactionFlow::class.qualifiedName!! )
         }
-
     }
 
     /**
@@ -384,6 +381,23 @@ class StaffedFlowHospital(private val flowMessaging: FlowMessaging, private val
             return exception.mentionsThrowable(SQLTransientConnectionException::class.java, "connection is not available")
         }
     }
+
+    /**
+     * Hospitalise any database (SQL) exception that wasn't handled otherwise, unless on the configurable whitelist
+     * Note that retry decisions from other specialists will not be affected as retries take precedence over hospitalisation.
+     */
+    object DatabaseEndocrinologist : Staff {
+        override fun consult(flowFiber: FlowFiber, currentState: StateMachineState, newError: Throwable, history: FlowMedicalHistory): Diagnosis {
+            return if (( newError is SQLException || newError is PersistenceException )&& !customConditions.any{it(newError)}) {
+                Diagnosis.OVERNIGHT_OBSERVATION
+            } else {
+                Diagnosis.NOT_MY_SPECIALTY
+            }
+        }
+
+        @VisibleForTesting
+        val customConditions = mutableSetOf<(t: Throwable) -> Boolean>()
+    }
 }
 
 private fun <T : Throwable> Throwable?.mentionsThrowable(exceptionType: Class<T>, errorMessage: String? = null): Boolean {
@@ -397,3 +411,4 @@ private fun <T : Throwable> Throwable?.mentionsThrowable(exceptionType: Class<T>
     }
     return (exceptionType.isAssignableFrom(this::class.java) && containsMessage) || cause.mentionsThrowable(exceptionType, errorMessage)
 }
+

node/src/test/kotlin/net/corda/node/services/statemachine/RetryFlowMockTest.kt

@@ -49,6 +49,8 @@ class RetryFlowMockTest {
         SendAndRetryFlow.count = 0
         RetryInsertFlow.count = 0
         KeepSendingFlow.count.set(0)
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.add { t -> t is LimitedRetryCausingError }
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.add { t -> t is RetryCausingError }
     }
 
     private fun <T> TestStartedNode.startFlow(logic: FlowLogic<T>): CordaFuture<T> {
@@ -58,6 +60,7 @@ class RetryFlowMockTest {
     @After
     fun cleanUp() {
         mockNet.stopNodes()
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.clear()
     }
 
     @Test

node/src/test/kotlin/net/corda/node/services/vault/VaultFlowTest.kt

@@ -6,6 +6,7 @@ import net.corda.core.identity.CordaX500Name
 import net.corda.core.identity.Party
 import net.corda.core.node.services.queryBy
 import net.corda.core.transactions.TransactionBuilder
+import net.corda.node.services.statemachine.StaffedFlowHospital
 import net.corda.testing.core.DummyCommandData
 import net.corda.testing.core.singleIdentity
 import net.corda.testing.internal.vault.DUMMY_DEAL_PROGRAM_ID
@@ -48,10 +49,12 @@ class VaultFlowTest {
     @After
     fun tearDown() {
         mockNetwork.stopNodes()
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.clear()
     }
 
     @Test
     fun `Unique column constraint failing causes states to not persist to vaults`() {
+        StaffedFlowHospital.DatabaseEndocrinologist.customConditions.add( { t: Throwable -> t is javax.persistence.PersistenceException })
         partyA.startFlow(Initiator(listOf(partyA.info.singleIdentity(), partyB.info.singleIdentity()))).get()
         Assertions.assertThatExceptionOfType(ExecutionException::class.java).isThrownBy {
             partyA.startFlow(Initiator(listOf(partyA.info.singleIdentity(), partyB.info.singleIdentity()))).get()