{185361198} tests: verify revconn skips rtcpu'd/decomm'd hosts

Add physrep_revconn_rtcpu to phys_rep_tiered.test, modeled on the existing
physrep_rtcpu_source case. Using the 'fakedr' trap to force machine_is_up()
to report a host as down, it asserts that the new 'revconn_host_is_up'
predicate (physrep_revconn_host_is_up) returns "up" for a healthy host,
"down" once the host is marked rtcpu'd, and "up" again after the mark is
cleared -- i.e. the reverse-connection sender and receiver will skip a
decommissioned/rtcpu'd host.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Signed-off-by: Mark Hannum <mhannum@bloomberg.net>

Author: markhannum

tests/phys_rep_tiered.test/runit

@@ -2649,6 +2649,53 @@ function physrep_rtcpu_source
     done
 }
 
+function physrep_revconn_rtcpu
+{
+    local out
+    local cmd
+    local sql
+    local rc
+
+    echo "== physrep_revconn_rtcpu =="
+
+    # Verify that the reverse-connection rtcpu predicate
+    # (physrep_revconn_host_is_up) used by both the reverse-conn sender
+    # (reverse_conn.c) and receiver (reversesql.c) treats an rtcpu'd /
+    # decommissioned host as "down", so we don't spin trying to establish
+    # reverse connections to a machine that is never coming back (DRQS
+    # 185361198). Exercised via the 'revconn_host_is_up' message command on the
+    # intermediary source db, which is the side that runs the revconn workers.
+    #
+    # Connections use --admin: $firstNode is itself one of $CLUSTER, so on the
+    # iteration where it is the target it gets marked rtcpu'd (drtest) and would
+    # otherwise refuse normal client traffic.
+    for tgt in $CLUSTER ; do
+        cmd="revconn_host_is_up $tgt"
+        sql="exec procedure sys.cmd.send(\"$cmd\")"
+
+        # Host is up before it is marked rtcpu'd -> connection allowed
+        out=$($CDB2SQL_EXE --tabs --admin $CDB2_OPTIONS ${REPL_CLUS_DBNAME} --host $firstNode "$sql")
+        rc=${out##*=}
+        [[ "$rc" == "1" ]] || cleanFailExit "revconn_host_is_up should allow $tgt before rtcpu, got rc=$rc"
+
+        # Mark the host rtcpu'd -> connection must be skipped
+        echo "Marking $tgt as RTCPU'd"
+        $CDB2SQL_EXE --admin $CDB2_OPTIONS ${REPL_CLUS_DBNAME} --host $firstNode \
+            "exec procedure sys.cmd.send(\"fakedr add $tgt\")"
+        out=$($CDB2SQL_EXE --tabs --admin $CDB2_OPTIONS ${REPL_CLUS_DBNAME} --host $firstNode "$sql")
+        rc=${out##*=}
+        [[ "$rc" == "0" ]] || cleanFailExit "revconn_host_is_up should skip rtcpu'd $tgt, got rc=$rc"
+
+        # Clear the rtcpu mark -> connection allowed again
+        echo "Removing RTCPU mark for $tgt"
+        $CDB2SQL_EXE --admin $CDB2_OPTIONS ${REPL_CLUS_DBNAME} --host $firstNode \
+            "exec procedure sys.cmd.send(\"fakedr del $tgt\")"
+        out=$($CDB2SQL_EXE --tabs --admin $CDB2_OPTIONS ${REPL_CLUS_DBNAME} --host $firstNode "$sql")
+        rc=${out##*=}
+        [[ "$rc" == "1" ]] || cleanFailExit "revconn_host_is_up should allow $tgt after removing rtcpu, got rc=$rc"
+    done
+}
+
 function physrep_block_writes_correct_rcode
 {
     # Make sure that physreps block writes & return the correct rcode when they do
@@ -3361,6 +3408,11 @@ function run_tests
     physrep_rtcpu_source
     testcase_finish $testcase
 
+    testcase="physrep_revconn_rtcpu"
+    testcase_preamble $testcase
+    physrep_revconn_rtcpu
+    testcase_finish $testcase
+
     testcase="test_intermediate_physrep_active_status"
     testcase_preamble $testcase
     test_intermediate_physrep_active_status