Orchagent changes needed to support single ASIC VOQ Fixed-System (#3847)

What I did

Portorch, Neighorch, Intfsorch are updated to not access chassis app DB. Chassis app DB is accessed only if the chassisdb.conf is present indicating its a real chassis (not a FS VOQ)
For lag creation, system lags are created as the switch is a VOQ swtich.
Why I did it

For a fixed system, there is no chassis DB present
How I verified it

Ran sonic-mgmt tests to verify BGP, LAG, functionality

Author: lakshmi-nexthop

.azure-pipelines/docker-sonic-vs/start.sh

@@ -61,10 +61,15 @@ else
         qos_cmd="-j /tmp/qos.json"
     fi
 
+    if [ -f /usr/share/sonic/single_asic_voq_fs/default_config.json ]; then
+        sonic-cfggen -j /usr/share/sonic/single_asic_voq_fs/default_config.json --print-data > /tmp/voq.json
+        voq_cmd="-j /tmp/voq.json"
+    fi
+
     sonic-cfggen -p /usr/share/sonic/device/$PLATFORM/$PLATFORM_CONF -k $HWSKU --print-data > /tmp/ports.json
     # change admin_status from up to down; Test cases dependent
     sed -i "s/up/down/g" /tmp/ports.json
-    sonic-cfggen -j /etc/sonic/init_cfg.json $buffers_cmd $qos_cmd -j /tmp/ports.json --print-data > /etc/sonic/config_db.json
+    sonic-cfggen -j /etc/sonic/init_cfg.json $buffers_cmd $qos_cmd $voq_cmd -j /tmp/ports.json --print-data > /etc/sonic/config_db.json
 fi
 
 sonic-cfggen -t /usr/share/sonic/templates/copp_cfg.j2 > /etc/sonic/copp_cfg.json

.azure-pipelines/test-docker-sonic-vs-template.yml

@@ -180,6 +180,8 @@ jobs:
       retry=3
       IMAGE_NAME=docker-sonic-vs:$(Build.DefinitionName).$(Build.BuildNumber).asan-${{ parameters.asan }}
       echo $all_tests | xargs -n 1 | xargs -P 8 -I TEST_MODULE sudo DEFAULT_CONTAINER_REGISTRY=publicmirror.azurecr.io/ ./run-tests.sh "$IMAGE_NAME" "$params" "TEST_MODULE" 3
+      single_asic_voq_tests="test_portchannel.py test_neighbor.py test_route.py"
+      echo $single_asic_voq_tests | xargs -n 1 | xargs -P 3 -I TEST_MODULE sudo ./run-tests.sh "$IMAGE_NAME" "--force-recreate-dvs --switch-mode=single_asic_voq_fs" "TEST_MODULE" 3
 
       rm -rf $(Build.ArtifactStagingDirectory)/download
     displayName: "Run vs tests"

orchagent/intfsorch.cpp

@@ -38,6 +38,7 @@ extern NeighOrch *gNeighOrch;
 extern string gMySwitchType;
 extern int32_t gVoqMySwitchId;
 extern bool gTraditionalFlexCounter;
+extern bool isChassisDbInUse();
 
 const int intfsorch_pri = 35;
 
@@ -98,7 +99,7 @@ IntfsOrch::IntfsOrch(DBConnector *db, string tableName, VRFOrch *vrf_orch, DBCon
                                  RIF_PLUGIN_FIELD,
                                  rifRateSha);
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         //Add subscriber to process VOQ system interface
         tableName = CHASSIS_APP_SYSTEM_INTERFACE_TABLE_NAME;
@@ -1310,7 +1311,7 @@ bool IntfsOrch::addRouterIntfs(sai_object_id_t vrf_id, Port &port, string loopba
 
     SWSS_LOG_NOTICE("Create router interface %s MTU %u", port.m_alias.c_str(), port.m_mtu);
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         // Sync the interface of local port/LAG to the SYSTEM_INTERFACE table of CHASSIS_APP_DB
         voqSyncAddIntf(port.m_alias);
@@ -1363,7 +1364,7 @@ bool IntfsOrch::removeRouterIntfs(Port &port)
 
     SWSS_LOG_NOTICE("Remove router interface for port %s", port.m_alias.c_str());
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         // Sync the removal of interface of local port/LAG to the SYSTEM_INTERFACE table of CHASSIS_APP_DB
         voqSyncDelIntf(port.m_alias);

orchagent/main.cpp

@@ -17,6 +17,8 @@ extern "C" {
 #include <stdexcept>
 #include <stdlib.h>
 #include <string.h>
+#include <fstream>
+#include <nlohmann/json.hpp>
 
 #include <sys/time.h>
 #include <sairedis.h>
@@ -77,6 +79,12 @@ string gMyHostName = "";
 string gMyAsicName = "";
 bool gTraditionalFlexCounter = false;
 uint32_t create_switch_timeout = 0;
+bool gMultiAsicVoq = false;
+
+bool isChassisDbInUse()
+{
+    return gMultiAsicVoq;
+}
 
 void usage()
 {
@@ -208,6 +216,18 @@ void getCfgSwitchType(DBConnector *cfgDb, string &switch_type, string &switch_su
 
 }
 
+bool isChassisAppDbPresent()
+{
+    std::ifstream file("/etc/sonic/database_config.json");
+    if (!file.is_open()) return false;
+
+    nlohmann::json db_config;
+    file >> db_config;
+
+    return db_config.contains("DATABASES") &&
+           db_config["DATABASES"].contains("CHASSIS_APP_DB");
+}
+
 bool getSystemPortConfigList(DBConnector *cfgDb, DBConnector *appDb, vector<sai_system_port_config_t> &sysportcfglist)
 {
     Table cfgDeviceMetaDataTable(cfgDb, CFG_DEVICE_METADATA_TABLE_NAME);
@@ -617,7 +637,19 @@ int main(int argc, char **argv)
 
         //Connect to CHASSIS_APP_DB in redis-server in control/supervisor card as per
         //connection info in database_config.json
-        chassis_app_db = make_shared<DBConnector>("CHASSIS_APP_DB", 0, true);
+        chassis_app_db = nullptr;
+        if (isChassisAppDbPresent())
+       	{
+            gMultiAsicVoq = true;
+            try
+            {
+                chassis_app_db = make_shared<DBConnector>("CHASSIS_APP_DB", 0, true);
+            }
+            catch (const std::exception& e)
+            {
+                SWSS_LOG_NOTICE("CHASSIS_APP_DB not available, operating in standalone VOQ mode");
+            }
+        }
     }
     else if (gMySwitchType == "fabric")
     {
@@ -867,6 +899,11 @@ int main(int argc, char **argv)
     }
 
     shared_ptr<OrchDaemon> orchDaemon;
+    DBConnector *chassis_db = nullptr;
+    if (chassis_app_db != nullptr)
+    {
+        chassis_db = chassis_app_db.get();
+    }
 
     /*
      * Declare shared pointers for dpu specific databases.
@@ -884,7 +921,7 @@ int main(int argc, char **argv)
 
     else if (gMySwitchType != "fabric")
     {
-        orchDaemon = make_shared<OrchDaemon>(&appl_db, &config_db, &state_db, chassis_app_db.get(), zmq_server.get());
+        orchDaemon = make_shared<OrchDaemon>(&appl_db, &config_db, &state_db, chassis_db, zmq_server.get());
         if (gMySwitchType == "voq")
         {
             orchDaemon->setFabricEnabled(true);
@@ -894,7 +931,7 @@ int main(int argc, char **argv)
     }
     else
     {
-        orchDaemon = make_shared<FabricOrchDaemon>(&appl_db, &config_db, &state_db, chassis_app_db.get(), zmq_server.get());
+        orchDaemon = make_shared<FabricOrchDaemon>(&appl_db, &config_db, &state_db, chassis_db, zmq_server.get());
     }
 
     if (gRingMode) {

orchagent/neighorch.cpp

@@ -25,6 +25,8 @@ extern BfdOrch *gBfdOrch;
 extern size_t gMaxBulkSize;
 extern string gMyHostName;
 
+extern bool isChassisDbInUse();
+
 const int neighorch_pri = 30;
 
 NeighOrch::NeighOrch(DBConnector *appDb, string tableName, IntfsOrch *intfsOrch, FdbOrch *fdbOrch, PortsOrch *portsOrch, DBConnector *chassisAppDb) :
@@ -46,7 +48,7 @@ NeighOrch::NeighOrch(DBConnector *appDb, string tableName, IntfsOrch *intfsOrch,
         gBfdOrch->attach(this);
     }
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         //Add subscriber to process VOQ system neigh
         tableName = CHASSIS_APP_SYSTEM_NEIGH_TABLE_NAME;
@@ -1230,7 +1232,7 @@ bool NeighOrch::addNeighbor(NeighborContext& ctx)
     NeighborUpdate update = { neighborEntry, macAddress, true };
     notify(SUBJECT_TYPE_NEIGH_CHANGE, static_cast<void *>(&update));
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         //Sync the neighbor to add to the CHASSIS_APP_DB
         voqSyncAddNeigh(alias, ip_address, macAddress, neighbor_entry);
@@ -1379,7 +1381,7 @@ bool NeighOrch::removeNeighbor(NeighborContext& ctx, bool disable)
     NeighborUpdate update = { neighborEntry, MacAddress(), false };
     notify(SUBJECT_TYPE_NEIGH_CHANGE, static_cast<void *>(&update));
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         //Sync the neighbor to delete from the CHASSIS_APP_DB
         voqSyncDelNeigh(alias, ip_address);

orchagent/p4orch/tests/test_main.cpp

@@ -38,6 +38,12 @@ sai_object_id_t gUnderlayIfId = 0x101;
 string gMyAsicName = "";
 event_handle_t g_events_handle;
 
+bool gMultiAsicVoq = false;
+bool isChassisDbInUse()
+{
+    return gMultiAsicVoq;
+}
+
 #define DEFAULT_BATCH_SIZE 128
 #define DEFAULT_MAX_BULK_SIZE 1000
 extern int gBatchSize;

orchagent/portsorch.cpp

@@ -69,6 +69,8 @@ extern int32_t gVoqMySwitchId;
 extern string gMyHostName;
 extern string gMyAsicName;
 extern event_handle_t g_events_handle;
+extern bool isChassisDbInUse();
+extern bool gMultiAsicVoq;
 
 // defines ------------------------------------------------------------------------------------------------------------
 
@@ -1002,7 +1004,7 @@ PortsOrch::PortsOrch(DBConnector *db, DBConnector *stateDb, vector<table_name_wi
         Orch::addExecutor(portHostTxReadyNotificatier);
     }
 
-    if (gMySwitchType == "voq")
+    if (isChassisDbInUse())
     {
         string tableName;
         //Add subscriber to process system LAG (System PortChannel) table
@@ -5998,7 +6000,7 @@ void PortsOrch::doLagMemberTask(Consumer &consumer)
                 }
             }
 
-            if ((gMySwitchType == "voq") && (port.m_type != Port::SYSTEM))
+            if (isChassisDbInUse() && (port.m_type != Port::SYSTEM))
             {
                //Sync to SYSTEM_LAG_MEMBER_TABLE of CHASSIS_APP_DB
                voqSyncAddLagMember(lag, port, status);
@@ -7616,13 +7618,16 @@ bool PortsOrch::addLag(string lag_alias, uint32_t spa_id, int32_t switch_id)
             switch_id = gVoqMySwitchId;
             system_lag_alias = gMyHostName + "|" + gMyAsicName + "|" + lag_alias;
 
-            // Allocate unique lag id
-            spa_id = m_lagIdAllocator->lagIdAdd(system_lag_alias, 0);
-
-            if ((int32_t)spa_id <= 0)
+            if (gMultiAsicVoq)
             {
-                SWSS_LOG_ERROR("Failed to allocate unique LAG id for local lag %s rv:%d", lag_alias.c_str(), spa_id);
-                return false;
+                // Allocate unique lag id
+                spa_id = m_lagIdAllocator->lagIdAdd(system_lag_alias, 0);
+
+                if ((int32_t)spa_id <= 0)
+                {
+                    SWSS_LOG_ERROR("Failed to allocate unique LAG id for local lag %s rv:%d", lag_alias.c_str(), spa_id);
+                    return false;
+                }
             }
         }
 
@@ -7734,7 +7739,7 @@ bool PortsOrch::removeLag(Port lag)
 
     m_counterLagTable->hdel("", lag.m_alias);
 
-    if (gMySwitchType == "voq")
+    if (isChassisDbInUse())
     {
         // Free the lag id, if this is local LAG
 
@@ -7847,7 +7852,7 @@ bool PortsOrch::addLagMember(Port &lag, Port &port, string member_status)
     LagMemberUpdate update = { lag, port, true };
     notify(SUBJECT_TYPE_LAG_MEMBER_CHANGE, static_cast<void *>(&update));
 
-    if (gMySwitchType == "voq")
+    if (isChassisDbInUse())
     {
         //Sync to SYSTEM_LAG_MEMBER_TABLE of CHASSIS_APP_DB
         voqSyncAddLagMember(lag, port, member_status);
@@ -7895,7 +7900,7 @@ bool PortsOrch::removeLagMember(Port &lag, Port &port)
     LagMemberUpdate update = { lag, port, false };
     notify(SUBJECT_TYPE_LAG_MEMBER_CHANGE, static_cast<void *>(&update));
 
-    if (gMySwitchType == "voq")
+    if (isChassisDbInUse())
     {
         //Sync to SYSTEM_LAG_MEMBER_TABLE of CHASSIS_APP_DB
         voqSyncDelLagMember(lag, port);
@@ -9132,7 +9137,7 @@ void PortsOrch::updatePortOperStatus(Port &port, sai_port_oper_status_t status)
         }
     }
 
-    if(gMySwitchType == "voq")
+    if(isChassisDbInUse())
     {
         if (gIntfsOrch->isLocalSystemPortIntf(port.m_alias))
         {
@@ -10279,7 +10284,8 @@ void PortsOrch::voqSyncAddLag (Port &lag)
 
     // Sync only local lag add to CHASSIS_APP_DB
 
-    if (switch_id != gVoqMySwitchId)
+    if (switch_id != gVoqMySwitchId ||
+       !gMultiAsicVoq)
     {
         return;
     }

tests/conftest.py

@@ -44,6 +44,8 @@
 # Voq asics will have 16 fabric ports created (defined in Azure/sonic-buildimage#7629).
 FABRIC_NUM_PORTS = 16
 
+SINGLE_ASIC_VOQ_FS = "single_asic_voq_fs"
+
 def ensure_system(cmd):
     rc, output = subprocess.getstatusoutput(cmd)
     if rc:
@@ -112,6 +114,12 @@ def pytest_addoption(parser):
                      default=False,
                      help="Collect the test coverage information")
 
+    parser.addoption("--switch-mode",
+                     action="store",
+                     default=None,
+                     type=str,
+                     help="Set switch mode information")
+
 
 def random_string(size=4, chars=string.ascii_uppercase + string.digits):
     return "".join(random.choice(chars) for x in range(size))
@@ -293,7 +301,8 @@ def __init__(
         newctnname: str = None,
         ctnmounts: Dict[str, str] = None,
         buffer_model: str = None,
-        enable_coverage: bool = False
+        enable_coverage: bool = False,
+        switch_mode: str = None
     ):
         self.basicd = ["redis-server", "rsyslogd"]
         self.swssd = [
@@ -316,6 +325,7 @@ def __init__(
         self.vct = vct
         self.ctn = None
         self.enable_coverage = enable_coverage
+        self.switch_mode = switch_mode
 
         self.cleanup = not keeptb
 
@@ -577,7 +587,10 @@ def check_swss_ready(self, timeout: int = 300) -> None:
         self.get_config_db()
         metadata = self.config_db.get_entry('DEVICE_METADATA|localhost', '')
         if metadata.get('switch_type', 'npu') in ['voq', 'fabric']:
-            num_ports = NUM_PORTS + FABRIC_NUM_PORTS
+            if self.switch_mode and self.switch_mode == SINGLE_ASIC_VOQ_FS:
+                num_ports = NUM_PORTS
+            else:
+                num_ports = NUM_PORTS + FABRIC_NUM_PORTS
 
         # Verify that all ports have been initialized and configured
         app_db = self.get_app_db()
@@ -597,8 +610,9 @@ def _polling_function():
 
         # Verify that fabric ports are monitored in STATE_DB
         if metadata.get('switch_type', 'npu') in ['voq', 'fabric']:
-            self.get_state_db()
-            self.state_db.wait_for_n_keys("FABRIC_PORT_TABLE", FABRIC_NUM_PORTS)
+            if not self.switch_mode or (self.switch_mode and self.switch_mode != SINGLE_ASIC_VOQ_FS):
+                self.get_state_db()
+                self.state_db.wait_for_n_keys("FABRIC_PORT_TABLE", FABRIC_NUM_PORTS)
 
     def net_cleanup(self) -> None:
         """Clean up network, remove extra links."""
@@ -1669,6 +1683,11 @@ def create_vct_ctn(self, ctndir):
             vol = {}
             vol[chassis_config_dir] = {"bind": "/usr/share/sonic/virtual_chassis", "mode": "ro"}
 
+            # Mount database_config.json when connect_to_chassis_db is set to 1
+            if defcfg.get("connect_to_chassis_db") == 1:
+                database_config_file = cwd + "/virtual_chassis/database_config.json"
+                vol[database_config_file] = {"bind": "/etc/sonic/database_config.json", "mode": "ro"}
+
             # pass self.ns into the vs to be use for vs restarts by swss conftest.
             # connection to chassbr is setup by chassis_connect.py within the vs
             data = {}
@@ -1859,6 +1878,7 @@ def manage_dvs(request) -> str:
     force_recreate = request.config.getoption("--force-recreate-dvs")
     graceful_stop = request.config.getoption("--graceful-stop")
     enable_coverage = request.config.getoption("--enable-coverage")
+    switch_mode = request.config.getoption("--switch-mode")
 
     dvs = None
     curr_dvs_env = [] # lgtm[py/unused-local-variable]
@@ -1890,7 +1910,13 @@ def update_dvs(log_path, new_dvs_env=[]):
                 dvs.get_logs()
                 dvs.destroy()
 
-            dvs = DockerVirtualSwitch(name, imgname, keeptb, new_dvs_env, log_path, max_cpu, forcedvs, buffer_model = buffer_model, enable_coverage=enable_coverage)
+            vol = {}
+            if switch_mode and switch_mode == SINGLE_ASIC_VOQ_FS:
+                cwd = os.getcwd()
+                voq_configs = cwd + "/single_asic_voq_fs"
+                vol[voq_configs] = {"bind": "/usr/share/sonic/single_asic_voq_fs", "mode": "ro"}
+
+            dvs = DockerVirtualSwitch(name, imgname, keeptb, new_dvs_env, log_path, max_cpu, forcedvs, buffer_model = buffer_model, enable_coverage=enable_coverage, ctnmounts=vol, switch_mode=switch_mode)
 
             curr_dvs_env = new_dvs_env