[WIP] Default thin-client to enabled and add proxy connectivity-probe gate

sdk/cosmos/azure-cosmos-tests/src/test/java/com/azure/cosmos/CosmosDiagnosticsTest.java

@@ -1981,14 +1981,21 @@ private void validateChannelAcquisitionContext(CosmosDiagnostics diagnostics, bo
     private String generateHttp2OptedInUserAgentIfRequired(String userAgent) {
         // Mirrors RxDocumentClientImpl.addUserAgentSuffix + UserAgentContainer.setFeatureEnabledFlagsAsSuffix:
         // when HTTP/2 is enabled, the Http2 bit is set; when PING keepalive is also effectively enabled
-        // (kill-switch on AND positive interval), the Http2PingHealth bit is OR'd in.
+        // (kill-switch on AND positive interval), the Http2PingHealth bit is OR'd in. ThinClient is set when
+        // COSMOS.THINCLIENT_ENABLED is true (default true after Gateway V2 default enablement).
         // Tests here do not override Http2ConnectionConfig.setEnabled(...) so the per-client override branch
         // in addUserAgentSuffix is a no-op for this helper.
+        int featureValue = 0;
+        if (Configs.isThinClientEnabled()) {
+            featureValue |= UserAgentFeatureFlags.ThinClient.getValue();
+        }
         if (Configs.isHttp2Enabled()) {
-            int featureValue = UserAgentFeatureFlags.Http2.getValue();
+            featureValue |= UserAgentFeatureFlags.Http2.getValue();
             if (Configs.isHttp2PingHealthEnabled() && Configs.getHttp2PingIntervalInSeconds() > 0) {
                 featureValue |= UserAgentFeatureFlags.Http2PingHealth.getValue();
             }
+        }
+        if (featureValue != 0) {
             userAgent = userAgent + "|F" + Integer.toHexString(featureValue).toUpperCase(Locale.ROOT);
         }
 

sdk/cosmos/azure-cosmos-tests/src/test/java/com/azure/cosmos/UserAgentSuffixTest.java

@@ -120,12 +120,19 @@ private void validateUserAgentSuffix(String actualUserAgent, String expectedUser
 
         // Mirrors RxDocumentClientImpl.addUserAgentSuffix + UserAgentContainer.setFeatureEnabledFlagsAsSuffix:
         // when HTTP/2 is enabled, the Http2 bit is set; when PING keepalive is also effectively enabled
-        // (kill-switch on AND positive interval), the Http2PingHealth bit is OR'd in.
+        // (kill-switch on AND positive interval), the Http2PingHealth bit is OR'd in. ThinClient is set when
+        // COSMOS.THINCLIENT_ENABLED is true (default true after Gateway V2 default enablement).
+        int featureValue = 0;
+        if (Configs.isThinClientEnabled()) {
+            featureValue |= UserAgentFeatureFlags.ThinClient.getValue();
+        }
         if (Configs.isHttp2Enabled()) {
-            int featureValue = UserAgentFeatureFlags.Http2.getValue();
+            featureValue |= UserAgentFeatureFlags.Http2.getValue();
             if (Configs.isHttp2PingHealthEnabled() && Configs.getHttp2PingIntervalInSeconds() > 0) {
                 featureValue |= UserAgentFeatureFlags.Http2PingHealth.getValue();
             }
+        }
+        if (featureValue != 0) {
             expectedUserAgentSuffix = expectedUserAgentSuffix + "|F" + Integer.toHexString(featureValue).toUpperCase(Locale.ROOT);
         }
 

sdk/cosmos/azure-cosmos-tests/src/test/java/com/azure/cosmos/faultinjection/Http2PingKeepaliveTest.java

@@ -86,6 +86,15 @@ public void beforeClass() {
         // set externally (Maven profile or -D) so a single test class covers both transports
         // across two pipeline runs.
         System.setProperty("COSMOS.HTTP2_ENABLED", "true");
+        // Disable the thin-client connectivity probe for the duration of this test.
+        // The probe fires HTTP/2 POSTs to thin-client port 10250 on every account refresh;
+        // when this test installs an iptables DROP on port 10250 (THIN_CLIENT_ENABLED=true
+        // branch) those probe requests time out, the probe trips
+        // isProxyProbeHealthy()->false, and routing falls back to Gateway V1 on port 443
+        // -- bypassing the very PING handler this test is exercising. Disable the probe
+        // so EndpointProbeClient.proxyHealthy stays optimistically true and the data plane
+        // request actually flows over port 10250 where the iptables DROP can take effect.
+        System.setProperty("COSMOS.THINCLIENT_PROBE_ENABLED", "false");
         logger.info("Transport selected: thinClient={}, h2Port={}", THIN_CLIENT_ENABLED, H2_PORT);
 
         this.client = getClientBuilder().buildAsyncClient();
@@ -101,6 +110,7 @@ public void beforeClass() {
     public void afterClass() {
         safeClose(this.client);
         System.clearProperty("COSMOS.HTTP2_ENABLED");
+        System.clearProperty("COSMOS.THINCLIENT_PROBE_ENABLED");
     }
 
     @BeforeMethod(groups = {"manual-http-network-fault"})

sdk/cosmos/azure-cosmos-tests/src/test/java/com/azure/cosmos/implementation/ClientConfigDiagnosticsTest.java

@@ -207,7 +207,10 @@ public void rntbd() throws Exception {
         assertThat(objectNode.get("connCfg").get("rntbd").asText()).isEqualTo("(cto:PT5S, nrto:PT5S, icto:PT0S, ieto:PT1H, mcpe:130, mrpc:30, cer:true)");
 
         String http2Enabled = Configs.isHttp2Enabled() ? "true" : "false";
-        assertThat(objectNode.get("connCfg").get("gw").asText()).isEqualTo("(cps:1000, nrto:PT1M, icto:PT1M, cto:PT45S, gwV2Cto:n/a, p:false, http2:(enabled:"+ http2Enabled + ", maxc:1000, minc:" + Math.max(8, Runtime.getRuntime().availableProcessors()) + ", maxs:30))");
+        String gwV2Cto = Configs.isThinClientEnabled()
+            ? Duration.ofMillis(Configs.getThinClientConnectionTimeoutInMs()).toString()
+            : "n/a";
+        assertThat(objectNode.get("connCfg").get("gw").asText()).isEqualTo("(cps:1000, nrto:PT1M, icto:PT1M, cto:PT45S, gwV2Cto:" + gwV2Cto + ", p:false, http2:(enabled:"+ http2Enabled + ", maxc:1000, minc:" + Math.max(8, Runtime.getRuntime().availableProcessors()) + ", maxs:30))");
         assertThat(objectNode.get("connCfg").get("other").asText()).isEqualTo("(ed: false, cs: false, rv: true)");
     }
 
@@ -244,7 +247,10 @@ public void gw() throws Exception {
         assertThat(objectNode.get("connCfg").get("rntbd").asText()).isEqualTo("null");
 
         String http2Enabled = Configs.isHttp2Enabled() ? "true" : "false";
-        assertThat(objectNode.get("connCfg").get("gw").asText()).isEqualTo("(cps:500, nrto:PT18S, icto:PT17S, cto:PT45S, gwV2Cto:n/a, p:false, http2:(enabled:" + http2Enabled + ", maxc:1000, minc:" + Math.max(8, Runtime.getRuntime().availableProcessors()) + ", maxs:30))");
+        String gwV2Cto = Configs.isThinClientEnabled()
+            ? Duration.ofMillis(Configs.getThinClientConnectionTimeoutInMs()).toString()
+            : "n/a";
+        assertThat(objectNode.get("connCfg").get("gw").asText()).isEqualTo("(cps:500, nrto:PT18S, icto:PT17S, cto:PT45S, gwV2Cto:" + gwV2Cto + ", p:false, http2:(enabled:" + http2Enabled + ", maxc:1000, minc:" + Math.max(8, Runtime.getRuntime().availableProcessors()) + ", maxs:30))");
         assertThat(objectNode.get("connCfg").get("other").asText()).isEqualTo("(ed: false, cs: false, rv: true)");
     }
 
@@ -318,7 +324,10 @@ public void full(
         assertThat(objectNode.get("connCfg").get("rntbd").asText()).isEqualTo("null");
 
         String http2Enabled = Configs.isHttp2Enabled() ? "true" : "false";
-        assertThat(objectNode.get("connCfg").get("gw").asText()).isEqualTo("(cps:500, nrto:PT18S, icto:PT17S, cto:PT45S, gwV2Cto:n/a, p:false, http2:(enabled:" + http2Enabled + ", maxc:1000, minc:" + Math.max(8, Runtime.getRuntime().availableProcessors()) + ", maxs:30))");
+        String gwV2Cto = Configs.isThinClientEnabled()
+            ? Duration.ofMillis(Configs.getThinClientConnectionTimeoutInMs()).toString()
+            : "n/a";
+        assertThat(objectNode.get("connCfg").get("gw").asText()).isEqualTo("(cps:500, nrto:PT18S, icto:PT17S, cto:PT45S, gwV2Cto:" + gwV2Cto + ", p:false, http2:(enabled:" + http2Enabled + ", maxc:1000, minc:" + Math.max(8, Runtime.getRuntime().availableProcessors()) + ", maxs:30))");
         assertThat(objectNode.get("connCfg").get("other").asText()).isEqualTo("(ed: true, cs: true, rv: false)");
         assertThat(objectNode.get("excrgns").asText()).isEqualTo("[westus2]");
 

sdk/cosmos/azure-cosmos-tests/src/test/java/com/azure/cosmos/implementation/ConfigsTests.java

@@ -232,8 +232,19 @@ public void http2MaxFrameSizeInBytes() {
 
     @Test(groups = { "emulator" })
     public void thinClientEnabledTest() {
-        assertThat(isThinClientEnabled()).isFalse();
+        // Default flipped to true in the probe-flow work; explicit "false" must still opt out.
         System.clearProperty("COSMOS.THINCLIENT_ENABLED");
+        try {
+            assertThat(isThinClientEnabled()).isTrue();
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_ENABLED");
+        }
+        System.setProperty("COSMOS.THINCLIENT_ENABLED", "false");
+        try {
+            assertThat(isThinClientEnabled()).isFalse();
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_ENABLED");
+        }
         System.setProperty("COSMOS.THINCLIENT_ENABLED", "true");
         try {
             assertThat(isThinClientEnabled()).isTrue();
@@ -242,6 +253,92 @@ public void thinClientEnabledTest() {
         }
     }
 
+    @Test(groups = { "unit" })
+    public void thinClientProbeEnabledDefaultTest() {
+        System.clearProperty("COSMOS.THINCLIENT_PROBE_ENABLED");
+        try {
+            assertThat(Configs.isThinClientProbeEnabled()).isTrue();
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_ENABLED");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbeEnabledOverrideTest() {
+        System.setProperty("COSMOS.THINCLIENT_PROBE_ENABLED", "false");
+        try {
+            assertThat(Configs.isThinClientProbeEnabled()).isFalse();
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_ENABLED");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbeFailureThresholdDefaultTest() {
+        System.clearProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD");
+        try {
+            assertThat(Configs.getThinClientProbeFailureThreshold()).isEqualTo(1);
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbeFailureThresholdOverrideTest() {
+        System.setProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD", "5");
+        try {
+            assertThat(Configs.getThinClientProbeFailureThreshold()).isEqualTo(5);
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbeFailureThresholdCoercionTest() {
+        System.setProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD", "0");
+        try {
+            assertThat(Configs.getThinClientProbeFailureThreshold()).isEqualTo(1);
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD");
+        }
+        System.setProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD", "-3");
+        try {
+            assertThat(Configs.getThinClientProbeFailureThreshold()).isEqualTo(1);
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbeFailureThresholdInvalidFallsBackToDefaultTest() {
+        System.setProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD", "not-a-number");
+        try {
+            assertThat(Configs.getThinClientProbeFailureThreshold()).isEqualTo(1);
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_FAILURE_THRESHOLD");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbePathDefaultTest() {
+        System.clearProperty("COSMOS.THINCLIENT_PROBE_PATH");
+        try {
+            assertThat(Configs.getThinClientProbePath()).isEqualTo("/connectivity-probe");
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_PATH");
+        }
+    }
+
+    @Test(groups = { "unit" })
+    public void thinClientProbePathOverrideTest() {
+        System.setProperty("COSMOS.THINCLIENT_PROBE_PATH", "/custom-probe");
+        try {
+            assertThat(Configs.getThinClientProbePath()).isEqualTo("/custom-probe");
+        } finally {
+            System.clearProperty("COSMOS.THINCLIENT_PROBE_PATH");
+        }
+    }
+
     @Test(groups = { "unit" })
     public void thinClientConnectionTimeoutDefaultTest() {
         // Default thin client connection timeout should be 5000 milliseconds