Fix stuck query after cancel or termination when segment is not responding

src/backend/cdb/cdbutil.c

@@ -662,6 +662,9 @@ cdbcomponent_cleanupIdleQEs(bool includeWriter)
 		}
 	}
 
+	/* reset flag in libpq, that is used to avoid stuck in a loop in PQcancel*/
+	PQbypassConnCloseAtCancel(false);
+
 	return;
 }
 

src/backend/fts/ftsprobe.c

@@ -29,6 +29,7 @@
 #include "postmaster/fts.h"
 #include "postmaster/ftsprobe.h"
 #include "postmaster/postmaster.h"
+#include "storage/pmsignal.h"
 #include "utils/snapmgr.h"
 
 
@@ -1175,6 +1176,7 @@ processResponse(fts_context *context)
 					   "triggered successfully",
 					   primary->config->segindex, primary->config->dbid);
 				ftsInfo->state = FTS_RESPONSE_PROCESSED;
+				SendPostmasterSignal(PMSIGNAL_FTS_PROMOTED_MIRROR);
 				break;
 			case FTS_SYNCREP_OFF_SUCCESS:
 				elogif(gp_log_fts >= GPVARS_VERBOSITY_VERBOSE, LOG,

src/backend/postmaster/postmaster.c

@@ -5690,6 +5690,29 @@ sigusr1_handler(SIGNAL_ARGS)
 		signal_child(DtxRecoveryPID(), SIGINT);
 	}
 
+	if (CheckPostmasterSignal(PMSIGNAL_FTS_PROMOTED_MIRROR))
+	{
+		/*
+		 * Notify all child coordinator backends that FTS has promoted a mirror.
+		 */
+		if (pmState == PM_RUN &&
+			GpIdentity.segindex == MASTER_CONTENT_ID)
+		{
+			dlist_iter	iter;
+			dlist_foreach(iter, &BackendList)
+			{
+				Backend *bp = dlist_container(Backend, elem, iter.cur);
+
+				if (bp->dead_end || !(BACKEND_TYPE_NORMAL & bp->bkend_type))
+					continue;
+
+				SendProcSignal(bp->pid,
+							   PROCSIG_FTS_PROMOTED_MIRROR,
+							   InvalidBackendId);
+			}
+		}
+	}
+
 	/*
 	 * Try to advance postmaster's state machine, if a child requests it.
 	 *

src/backend/storage/ipc/procsignal.c

@@ -19,6 +19,7 @@
 
 #include "cdb/cdbvars.h"
 #include "commands/async.h"
+#include "libpq-fe.h"
 #include "miscadmin.h"
 #include "replication/walsender.h"
 #include "storage/latch.h"
@@ -288,6 +289,23 @@ QueryFinishHandler(void)
 	}
 }
 
+/*
+ * Coordinator postmaster signals that FTS has detected a failed segment and 
+ * promoted the mirror.  
+ */
+static void
+FtsPromotedMirrorHandler(void)
+{
+	/*
+	 * In case the promotion is done during cancel or termination of a query,
+	 * there is a very high chance that libpq will be stuck forever trying to
+	 * wait cancel confirmation from the segment, which is not responding. Code
+	 * below handles this case.
+	 */
+	if (QueryCancelCleanup || TermSignalReceived)
+		PQbypassConnCloseAtCancel(true);
+}
+
 /*
  * procsignal_sigusr1_handler - handle SIGUSR1 signal.
  */
@@ -329,6 +347,9 @@ procsignal_sigusr1_handler(SIGNAL_ARGS)
 	if (CheckProcSignal(PROCSIG_RESOURCE_GROUP_MOVE_QUERY))
 		HandleMoveResourceGroup();
 
+	if (CheckProcSignal(PROCSIG_FTS_PROMOTED_MIRROR))
+		FtsPromotedMirrorHandler();
+
 	if (set_latch_on_sigusr1 && MyProc != NULL)
 		SetLatch(&MyProc->procLatch);
 

src/include/storage/pmsignal.h

@@ -35,6 +35,8 @@ typedef enum
 	PMSIGNAL_WAKEN_FTS,         /* wake up FTS to probe segments */
 	PMSIGNAL_WAKEN_DTX_RECOVERY,         /* wake up dtx recovery to abort dtx xacts */
 
+	PMSIGNAL_FTS_PROMOTED_MIRROR, /* FTS has detected failed primary and promoted mirror*/
+
 	NUM_PMSIGNALS				/* Must be last value of enum! */
 } PMSignalReason;
 

src/include/storage/procsignal.h

@@ -43,6 +43,9 @@ typedef enum
 	PROCSIG_RECOVERY_CONFLICT_BUFFERPIN,
 	PROCSIG_RECOVERY_CONFLICT_STARTUP_DEADLOCK,
 	PROCSIG_RESOURCE_GROUP_MOVE_QUERY,	/* move query to a new resource group */
+
+	PROCSIG_FTS_PROMOTED_MIRROR, /* FTS has detected failed primary and promoted mirror*/
+
 	NUM_PROCSIGNALS				/* Must be last! */
 } ProcSignalReason;
 

src/interfaces/libpq/fe-connect.c

@@ -391,6 +391,8 @@ static const PQEnvironmentOption EnvironmentOptions[] =
 static const char uri_designator[] = "postgresql://";
 static const char short_uri_designator[] = "postgres://";
 
+static bool bypass_conn_close_at_cancel = false;
+
 static bool connectOptions1(PGconn *conn, const char *conninfo);
 static bool connectOptions2(PGconn *conn);
 static int	connectDBStart(PGconn *conn);
@@ -3429,6 +3431,10 @@ PQfreeCancel(PGcancel *cancel)
 		free(cancel);
 }
 
+void PQbypassConnCloseAtCancel(pqbool bypass)
+{
+	bypass_conn_close_at_cancel = bypass;
+}
 
 /*
  * PQcancel and PQrequestCancel: attempt to request cancellation of the
@@ -3523,6 +3529,10 @@ internal_cancel(SockAddr *raddr, int be_pid, int be_key,
 	 */
 #ifndef WIN32
 retry5:
+
+	if (bypass_conn_close_at_cancel)
+		goto cancel_errReturn;
+
 	pollFds[0].fd = tmpsock;
 	pollFds[0].events = POLLIN;
 	pollFds[0].revents = 0;

src/interfaces/libpq/libpq-fe.h

@@ -325,6 +325,8 @@ extern void PQfreeCancel(PGcancel *cancel);
 /* issue a cancel request */
 extern int	PQcancel(PGcancel *cancel, char *errbuf, int errbufsize);
 
+extern void PQbypassConnCloseAtCancel(pqbool bypass); /* GPDB only */
+
 /* issue a finsh request */
 extern int	PQrequestFinish(PGcancel *cancel, char *errbuf, int errbufsize);