Add port flap count and last flap timestamp to APPL_DB (sonic-net#3052)

* Add port flap count and last flap timestamp
mssonicbld · Jul 31, 2024 · d6e75e2 · d6e75e2
1 parent 1cbaf9d
commit d6e75e2
Show file tree

Hide file tree

Showing 5 changed files with 117 additions and 3 deletions.
diff --git a/orchagent/port.h b/orchagent/port.h
@@ -164,6 +164,7 @@ class Port
     uint32_t  m_nat_zone_id = 0;
     uint32_t  m_vnid = VNID_NONE;
     uint32_t  m_fdb_count = 0;
+    uint64_t  m_flap_count = 0;
     uint32_t  m_up_member_count = 0;
     uint32_t  m_maximum_headroom = 0;
     std::set<uint32_t> m_adv_speeds;

diff --git a/orchagent/portsorch.cpp b/orchagent/portsorch.cpp
@@ -3097,6 +3097,30 @@ bool PortsOrch::removeVlanHostIntf(Port vl)
     return true;
 }
 
+void PortsOrch::updateDbPortFlapCount(Port& port, sai_port_oper_status_t pstatus)
+{
+    SWSS_LOG_ENTER();
+
+    ++port.m_flap_count;
+    vector<FieldValueTuple> tuples;
+    FieldValueTuple tuple("flap_count", std::to_string(port.m_flap_count));
+    tuples.push_back(tuple);
+
+    auto now = std::chrono::system_clock::now();
+    std::time_t now_c = std::chrono::system_clock::to_time_t(now);
+    if (pstatus == SAI_PORT_OPER_STATUS_DOWN)
+    {
+        FieldValueTuple tuple("last_down_time", std::ctime(&now_c));
+        tuples.push_back(tuple);
+    } 
+    else if (pstatus == SAI_PORT_OPER_STATUS_UP) 
+    {
+        FieldValueTuple tuple("last_up_time", std::ctime(&now_c));
+        tuples.push_back(tuple);
+    }
+    m_portTable->set(port.m_alias, tuples);
+}
+
 void PortsOrch::updateDbPortOperStatus(const Port& port, sai_port_oper_status_t status) const
 {
     SWSS_LOG_ENTER();
@@ -5354,7 +5378,7 @@ bool PortsOrch::initializePort(Port &port)
     /* Check warm start states */
     vector<FieldValueTuple> tuples;
     bool exist = m_portTable->get(port.m_alias, tuples);
-    string operStatus;
+    string operStatus, flapCount = "0";
     if (exist)
     {
         for (auto i : tuples)
@@ -5363,9 +5387,14 @@ bool PortsOrch::initializePort(Port &port)
             {
                 operStatus = fvValue(i);
             }
+
+            if (fvField(i) == "flap_count")
+            {
+                flapCount = fvValue(i);
+            }
         }
     }
-    SWSS_LOG_DEBUG("initializePort %s with oper %s", port.m_alias.c_str(), operStatus.c_str());
+    SWSS_LOG_INFO("Port %s with oper %s flap_count=%s", port.m_alias.c_str(), operStatus.c_str(), flapCount.c_str());
 
     /**
      * Create database port oper status as DOWN if attr missing
@@ -5386,6 +5415,20 @@ bool PortsOrch::initializePort(Port &port)
         port.m_oper_status = SAI_PORT_OPER_STATUS_DOWN;
     }
 
+    // initalize port flap count
+    if (!flapCount.empty())
+    {
+        try
+        {
+            port.m_flap_count = stoull(flapCount);
+            m_portTable->hset(port.m_alias, "flap_count", flapCount);
+        }
+        catch (const std::exception &e)
+        {
+            SWSS_LOG_ERROR("Failed to get port (%s) flap_count: %s", port.m_alias.c_str(), e.what());
+        }
+    }
+
     /* initialize port admin status */
     if (!getPortAdminStatus(port.m_port_id, port.m_admin_state_up))
     {
@@ -7625,6 +7668,7 @@ void PortsOrch::updatePortOperStatus(Port &port, sai_port_oper_status_t status)
     if (port.m_type == Port::PHY)
     {
         updateDbPortOperStatus(port, status);
+        updateDbPortFlapCount(port, status);
         updateGearboxPortOperStatus(port);
 
         /* Refresh the port states and reschedule the poller tasks */

diff --git a/orchagent/portsorch.h b/orchagent/portsorch.h
@@ -147,6 +147,7 @@ class PortsOrch : public Orch, public Subject
 
     bool setHostIntfsOperStatus(const Port& port, bool up) const;
     void updateDbPortOperStatus(const Port& port, sai_port_oper_status_t status) const;
+    void updateDbPortFlapCount(Port& port, sai_port_oper_status_t pstatus);
 
     bool createVlanHostIntf(Port& vl, string hostif_name);
     bool removeVlanHostIntf(Port vl);

diff --git a/tests/mock_tests/portsorch_ut.cpp b/tests/mock_tests/portsorch_ut.cpp
@@ -504,6 +504,73 @@ namespace portsorch_test
         }
 
     };
+
+    /*
+    * Test port flap count 
+    */
+    TEST_F(PortsOrchTest, PortFlapCount)
+    {
+        Table portTable = Table(m_app_db.get(), APP_PORT_TABLE_NAME);
+
+        // Get SAI default ports to populate DB
+        auto ports = ut_helper::getInitialSaiPorts();
+
+        // Populate port table with SAI ports
+        for (const auto &it : ports)
+        {
+            portTable.set(it.first, it.second);
+        }
+
+        // Set PortConfigDone, PortInitDone
+        portTable.set("PortConfigDone", { { "count", to_string(ports.size()) } });
+        portTable.set("PortInitDone", { { "lanes", "0" } });
+
+        // refill consumer
+        gPortsOrch->addExistingData(&portTable);
+        // Apply configuration : create ports
+        static_cast<Orch *>(gPortsOrch)->doTask();
+
+        // Get first port, expect the oper status is not UP
+        Port port;
+        gPortsOrch->getPort("Ethernet0", port);
+        ASSERT_TRUE(port.m_oper_status != SAI_PORT_OPER_STATUS_UP);
+        ASSERT_TRUE(port.m_flap_count == 0);
+
+        auto exec = static_cast<Notifier *>(gPortsOrch->getExecutor("PORT_STATUS_NOTIFICATIONS"));
+        auto consumer = exec->getNotificationConsumer();
+
+        // mock a redis reply for notification, it notifies that Ehernet0 is going to up
+        for (uint32_t count=0; count < 5; count++) {
+            sai_port_oper_status_t oper_status = (count % 2 == 0) ? SAI_PORT_OPER_STATUS_UP : SAI_PORT_OPER_STATUS_DOWN;
+            mockReply = (redisReply *)calloc(sizeof(redisReply), 1);
+            mockReply->type = REDIS_REPLY_ARRAY;
+            mockReply->elements = 3; // REDIS_PUBLISH_MESSAGE_ELEMNTS
+            mockReply->element = (redisReply **)calloc(sizeof(redisReply *), mockReply->elements);
+            mockReply->element[2] = (redisReply *)calloc(sizeof(redisReply), 1);
+            mockReply->element[2]->type = REDIS_REPLY_STRING;
+            sai_port_oper_status_notification_t port_oper_status;
+            port_oper_status.port_state = oper_status;
+            port_oper_status.port_id = port.m_port_id;
+            std::string data = sai_serialize_port_oper_status_ntf(1, &port_oper_status);
+            std::vector<FieldValueTuple> notifyValues;
+            FieldValueTuple opdata("port_state_change", data);
+            notifyValues.push_back(opdata);
+            std::string msg = swss::JSon::buildJson(notifyValues);
+            mockReply->element[2]->str = (char*)calloc(1, msg.length() + 1);
+            memcpy(mockReply->element[2]->str, msg.c_str(), msg.length());
+
+            // trigger the notification
+            consumer->readData();
+            gPortsOrch->doTask(*consumer);
+            mockReply = nullptr;
+
+            gPortsOrch->getPort("Ethernet0", port);
+            ASSERT_TRUE(port.m_oper_status == oper_status);
+            ASSERT_TRUE(port.m_flap_count == count+1);
+        }
+
+        cleanupPorts(gPortsOrch);
+    }
 
     TEST_F(PortsOrchTest, PortBulkCreateRemove)
     {
@@ -2038,6 +2105,7 @@ namespace portsorch_test
 
         gPortsOrch->getPort("Ethernet0", port);
         ASSERT_TRUE(port.m_oper_status == SAI_PORT_OPER_STATUS_UP);
+        ASSERT_TRUE(port.m_flap_count == 1);
 
         std::vector<FieldValueTuple> values;
         portTable.get("Ethernet0", values);

diff --git a/tests/test_warm_reboot.py b/tests/test_warm_reboot.py
@@ -1090,7 +1090,7 @@ def test_swss_port_state_syncup(self, dvs, testlog):
                     orchStateCount += 1;
 
         # Only WARM_RESTART_TABLE|orchagent state=reconciled operation may exist after port oper status change.
-        assert orchStateCount == 1
+        assert orchStateCount == 2
 
         #clean up arp
         dvs.runcmd("arp -d 10.0.0.1")