From: Aram Sargsyan <aram@isc.org>
Date: Thu, 6 Mar 2025 14:37:09 +0000 (+0000)
Subject: Test resolver statistics when responses time out
X-Git-Tag: v9.18.36~7^2
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=006130a35c219957b8a799e2c99f36ff746ad253;p=thirdparty%2Fbind9.git

Test resolver statistics when responses time out

Add a test to check that the timed out responses do not skew the
normal responses statistics counters, and that they do update the
timeouts counter.

(cherry picked from commit 0c7fa8d572bf3e742a627ff660175683e131908b)
---

diff --git a/bin/tests/system/resolver/ans3/ans.pl b/bin/tests/system/resolver/ans3/ans.pl
index 880848e9c29..f9690dda2ae 100644
--- a/bin/tests/system/resolver/ans3/ans.pl
+++ b/bin/tests/system/resolver/ans3/ans.pl
@@ -74,6 +74,10 @@ sub handleQuery {
 		$packet->push("answer", new Net::DNS::RR($qname . " 300 A 10.53.0.3"));
 	} elsif ($qname eq "nodata.example.net") {
 		# Do not add a SOA RRset.
+	} elsif ($qname eq "noresponse.example.net") {
+		# Do not response.
+		print "RESPONSE:\n";
+		return "";
 	} elsif ($qname eq "nxdomain.example.net") {
 		# Do not add a SOA RRset.
 		$packet->header->rcode(NXDOMAIN);
@@ -185,8 +189,12 @@ for (;;) {
 			print "TCP request\n";
 			my $result = handleQuery($buf);
 			$len = length($result);
-			$conn->syswrite(pack("n", $len), 2);
-			$n = $conn->syswrite($result, $len);
+			if ($len != 0) {
+				$conn->syswrite(pack("n", $len), 2);
+				$n = $conn->syswrite($result, $len);
+			} else {
+				$n = 0;
+			}
 			print "    Sent: $n chars via TCP\n";
 		}
 		$conn->close;
diff --git a/bin/tests/system/resolver/tests.sh b/bin/tests/system/resolver/tests.sh
index 23b42f728cd..2864aae9501 100755
--- a/bin/tests/system/resolver/tests.sh
+++ b/bin/tests/system/resolver/tests.sh
@@ -85,6 +85,37 @@ if [ -x "${RESOLVE}" ]; then
   status=$((status + ret))
 fi
 
+rndccmd 10.53.0.1 stats || ret=1 # Get the responses, RTT and timeout statistics before the following timeout tests
+grep -F 'responses received' ns1/named.stats >ns1/named.stats.responses-before || true
+grep -F 'queries with RTT' ns1/named.stats >ns1/named.stats.rtt-before || true
+grep -F 'query timeouts' ns1/named.stats >ns1/named.stats.timeouts-before || true
+mv ns1/named.stats ns1/named.stats-before
+
+# Checking if the "hung fetch" timer kicks in and interrupts a non-responsive query.
+n=$((n + 1))
+echo_i "checking no response timeout handling ($n)"
+ret=0
+dig_with_opts +tcp +tries=1 +timeout=15 noresponse.example.net @10.53.0.1 a >dig.out.ns1.test${n} || ret=1
+grep -F "status: SERVFAIL" dig.out.ns1.test${n} >/dev/null || ret=1
+if [ $ret != 0 ]; then echo_i "failed"; fi
+status=$((status + ret))
+
+n=$((n + 1))
+echo_i "checking that the timeout didn't skew the resolver responses counters and did update the timeout counter ($n)"
+ret=0
+rndccmd 10.53.0.1 stats || ret=1
+grep -F 'responses received' ns1/named.stats >ns1/named.stats.responses-after || true
+grep -F 'queries with RTT' ns1/named.stats >ns1/named.stats.rtt-after || true
+grep -F 'query timeouts' ns1/named.stats >ns1/named.stats.timeouts-after || true
+mv ns1/named.stats ns1/named.stats-after
+diff ns1/named.stats.responses-before ns1/named.stats.responses-after >/dev/null || ret=1
+diff ns1/named.stats.rtt-before ns1/named.stats.rtt-after >/dev/null || ret=1
+# The following check is disabled in this branch, because TCP timeouts don't
+# work well here, and instead the "hung fetch" timer interrupts the query.
+#diff ns1/named.stats.timeouts-before ns1/named.stats.timeouts-after >/dev/null && ret=1
+if [ $ret != 0 ]; then echo_i "failed"; fi
+status=$((status + ret))
+
 n=$((n + 1))
 echo_i "checking handling of bogus referrals ($n)"
 # If the server has the "INSIST(!external)" bug, this query will kill it.
diff --git a/bin/tests/system/resolver/tests_sh_resolver.py b/bin/tests/system/resolver/tests_sh_resolver.py
index 4677a8337e0..2362ae17260 100644
--- a/bin/tests/system/resolver/tests_sh_resolver.py
+++ b/bin/tests/system/resolver/tests_sh_resolver.py
@@ -22,6 +22,7 @@ pytestmark = pytest.mark.extra_artifacts(
         "resolve.out.*",
         "ans*/ans.run",
         "ans*/query.log",
+        "ns1/named.stats*",
         "ns4/tld.db",
         "ns5/trusted.conf",
         "ns6/K*",