WL#9638: Step 39

----------------
BUG#27942583
COPY_FRAGREQ scans can be blocked on row locks without
any real time limit. So continue writing node log messages,
but don't crash because this scan is taking a long time.
Node restart will take an extreme amount of time, but we
should not kill the live node because of that. If any node
is to be killed, it is rather the starting node.
Will still crash on too long waits for LCP scans and Backup scans
since those ignore locks. Not sure though if there is any real
possibility for those to be stalled for that long.
This check was introduced in 7.6, so this bug should not affect
any GA release.

(cherry picked from commit 43bd5be81a3ed894ead380812f263bae48844d2f)

Author: Mikael Ronström

storage/ndb/src/kernel/blocks/dblqh/DblqhMain.cpp

@@ -30428,7 +30428,7 @@ Dblqh::handle_check_system_scans(Signal *signal)
         else if (i == ZBACKUP_CHECK_INDEX)
         {
           jam();
-          g_eventLogger->info("Backup Scan have stalled for %u seconds, last"
+          g_eventLogger->info("Backup Scan have stalled for %u seconds, "
                               "last seen on line %u, check_lcp_stop_count: %u",
                               time_stalled,
                               c_check_scanptr_save_line[i],
@@ -30449,8 +30449,15 @@ Dblqh::handle_check_system_scans(Signal *signal)
         signal->theData[0] = DumpStateOrd::LqhDumpOneScanRec;
         signal->theData[1] = loc_scanptr.i;
         EXECUTE_DIRECT(DBLQH, GSN_DUMP_STATE_ORD, signal, 2);
-        if (time_stalled >= 120)
+        if (time_stalled >= 120 &&
+            (i != ZCOPY_FRAGREQ_CHECK_INDEX))
         {
+          /**
+           * LCP and Backup scans proceed even in the presence of locks,
+           * COPY_FRAGREQ scans can be held up on locks without any
+           * real limits since transactions have no specified maximum
+           * time.
+           */
           jam();
           abort();
         }