[#24658] YSQL: Switch pg_cron to normal SIGTERM

Summary:
Switch pg_cron to normal using normal `SIGTERM` and `SIGQUIT` handlers.

In an attempt to fix stuck pg_cron workers the handler of `SIGTERM` was changed to `quickdie` 4b4c201/D37591. But in `quickdie` mode postgres does not invoke `ReportBackgroundWorkerExit` which signals the parent process when its child backend exits. This causes pg_cron launcher to get stuck in `WaitForBackgroundWorkerShutdown`.

With 111b65d/D39207 the issues with pg shutdown have been addressed and it is safe to now switch to the normal SIGTERM behavior which makes the background workers `die` in a clean manner.

#24706 tracks the issue where in yb `ReportBackgroundWorkerExit` is not called when the backend has a non graceful exit.

Fixes #24658
Jira: DB-13724

Test Plan: PgCronTest, DeactivateRunningJob

Reviewers: telgersma

Reviewed By: telgersma

Subscribers: ybase, yql

Differential Revision: https://phorge.dev.yugabyte.com/D39552

src/postgres/third-party-extensions/pg_cron/src/pg_cron.c

@@ -595,9 +595,7 @@ PgCronLauncherMain(Datum arg)
 	/* Establish signal handlers before unblocking signals. */
 	pqsignal(SIGHUP, pg_cron_sighup);
 	pqsignal(SIGINT, SIG_IGN);
-	/* YB Note: Exit immediately. */
-	pqsignal(SIGTERM, quickdie);
-	pqsignal(SIGQUIT, quickdie);
+	pqsignal(SIGTERM, pg_cron_sigterm);
 
 	/* We're now ready to receive signals */
 	BackgroundWorkerUnblockSignals();
@@ -2157,9 +2155,7 @@ CronBackgroundWorker(Datum main_arg)
 	shm_mq_handle *responseq;
 
 	/* handle SIGTERM like regular backend */
-	pqsignal(SIGTERM, quickdie);
-	/* YB Note: Exit immediately. */
-	pqsignal(SIGQUIT, quickdie);
+	pqsignal(SIGTERM, die);
 	BackgroundWorkerUnblockSignals();
 
 	/* Set up a memory context and resource owner. */

src/yb/integration-tests/pg_cron-test.cc

@@ -145,6 +145,15 @@ class PgCronTest : public MiniClusterTestWithClient<ExternalMiniCluster> {
     return conn->FetchRow<pgwrapper::PGUint64>(Format("SELECT COUNT(*) FROM $0", table_name));
   }
 
+  Status WaitForRowCountAbove(int64_t min_count) {
+    return WaitFor(
+        [this, &min_count]() -> Result<bool> {
+          const auto row_count = VERIFY_RESULT(GetRowCount());
+          return row_count > min_count;
+        },
+        kTimeout, Format("Wait for row count to be above $0", min_count));
+  }
+
   Status WaitForDataInPgCronLeaderTable() {
     auto table_name =
         stateful_service::GetStatefulServiceTableName(StatefulServiceKind::PG_CRON_LEADER);
@@ -542,4 +551,35 @@ TEST_F(PgCronTest, FailBeforeStoringLastMinute) {
   ASSERT_GT(row_count, 0);
 }
 
+TEST_F(PgCronTest, DeactivateRunningJob) {
+  ASSERT_OK(Schedule1SecInsertJob());
+  // Start a job that will run for a long time.
+  auto sleep_job_id = ASSERT_RESULT(ScheduleJob("Sleep Job", "1 second", "SELECT pg_sleep(1000)"));
+
+  ASSERT_OK(WaitForRowCountAbove(0));
+
+  auto is_sleep_job_running = [this, sleep_job_id]() -> Result<bool> {
+    return VERIFY_RESULT(conn_->FetchRow<pgwrapper::PGUint64>(Format(
+               "SELECT COUNT(*) FROM cron.job_run_details WHERE jobid = $0 AND status = 'running'",
+               sleep_job_id))) > 0;
+  };
+
+  {
+    const auto is_running = ASSERT_RESULT(is_sleep_job_running());
+    ASSERT_TRUE(is_running);
+  }
+
+  // Deactivating the sleep job should stop it immediately.
+  ASSERT_OK(conn_->FetchFormat("SELECT cron.alter_job($0, active:=false)", sleep_job_id));
+
+  // Wait for the change to get picked up and sleep job to get canceled.
+  ASSERT_OK(WaitFor(
+      [&is_sleep_job_running]() -> Result<bool> { return !VERIFY_RESULT(is_sleep_job_running()); },
+      kTimeout, "Wait for sleeping job to get killed"));
+
+  // Make sure the other job is still running.
+  const auto initial_row_count = ASSERT_RESULT(GetRowCount());
+  ASSERT_OK(WaitForRowCountAbove(initial_row_count));
+}
+
 }  // namespace yb