[Evpn Warmreboot] Added Dependancy check logic in VrfMgr (sonic-net#1466)

* [Evpn Warmreboot] Added Dependancy check logic in VrfMgr
This was done to ensure for EVPN warm-reboot the order of data replay to kernel is maintained across various submodules and the kernel programming will be successful.

* Marking Vrfmgrd and Intfmgrd to reconcile immediately after replay
There is no reconcile operation required in vrfmgrd and intfmgrd, hence immediately after replay these are marked as reconciled

Author: nkelapur

cfgmgr/intfmgr.cpp

@@ -24,6 +24,10 @@ using namespace swss;
 
 IntfMgr::IntfMgr(DBConnector *cfgDb, DBConnector *appDb, DBConnector *stateDb, const vector<string> &tableNames) :
         Orch(cfgDb, tableNames),
+        m_cfgIntfTable(cfgDb, CFG_INTF_TABLE_NAME),
+        m_cfgVlanIntfTable(cfgDb, CFG_VLAN_INTF_TABLE_NAME),
+        m_cfgLagIntfTable(cfgDb, CFG_LAG_INTF_TABLE_NAME),
+        m_cfgLoopbackIntfTable(cfgDb, CFG_LOOPBACK_INTERFACE_TABLE_NAME),
         m_statePortTable(stateDb, STATE_PORT_TABLE_NAME),
         m_stateLagTable(stateDb, STATE_LAG_TABLE_NAME),
         m_stateVlanTable(stateDb, STATE_VLAN_TABLE_NAME),
@@ -34,6 +38,12 @@ IntfMgr::IntfMgr(DBConnector *cfgDb, DBConnector *appDb, DBConnector *stateDb, c
     if (!WarmStart::isWarmStart())
     {
         flushLoopbackIntfs();
+        WarmStart::setWarmStartState("intfmgrd", WarmStart::WSDISABLED);
+    }
+    else
+    {
+        //Build the interface list to be replayed to Kernel
+        buildIntfReplayList();
     }
 }
 
@@ -172,6 +182,25 @@ int IntfMgr::getIntfIpCount(const string &alias)
     return std::stoi(res);
 }
 
+void IntfMgr::buildIntfReplayList(void)
+{
+    vector<string> intfList;
+
+    m_cfgIntfTable.getKeys(intfList);
+    std::copy( intfList.begin(), intfList.end(), std::inserter( m_pendingReplayIntfList, m_pendingReplayIntfList.end() ) );
+
+    m_cfgLoopbackIntfTable.getKeys(intfList);
+    std::copy( intfList.begin(), intfList.end(), std::inserter( m_pendingReplayIntfList, m_pendingReplayIntfList.end() ) );
+        
+    m_cfgVlanIntfTable.getKeys(intfList);
+    std::copy( intfList.begin(), intfList.end(), std::inserter( m_pendingReplayIntfList, m_pendingReplayIntfList.end() ) );
+        
+    m_cfgLagIntfTable.getKeys(intfList);
+    std::copy( intfList.begin(), intfList.end(), std::inserter( m_pendingReplayIntfList, m_pendingReplayIntfList.end() ) );
+
+    SWSS_LOG_INFO("Found %d Total Intfs to be replayed", (int)m_pendingReplayIntfList.size() );
+}
+
 bool IntfMgr::isIntfCreated(const string &alias)
 {
     vector<FieldValueTuple> temp;
@@ -676,6 +705,7 @@ bool IntfMgr::doIntfAddrTask(const vector<string>& keys,
 void IntfMgr::doTask(Consumer &consumer)
 {
     SWSS_LOG_ENTER();
+    static bool replayDone = false;
 
     auto it = consumer.m_toSync.begin();
     while (it != consumer.m_toSync.end())
@@ -693,6 +723,11 @@ void IntfMgr::doTask(Consumer &consumer)
                 it++;
                 continue;
             }
+            else
+            {
+                //Entry programmed, remove it from pending list if present
+                m_pendingReplayIntfList.erase(keys[0]);
+            }
         }
         else if (keys.size() == 2)
         {
@@ -701,6 +736,11 @@ void IntfMgr::doTask(Consumer &consumer)
                 it++;
                 continue;
             }
+            else
+            {
+                //Entry programmed, remove it from pending list if present
+                m_pendingReplayIntfList.erase(keys[0] + config_db_key_delimiter + keys[1] );
+            }
         }
         else
         {
@@ -709,4 +749,13 @@ void IntfMgr::doTask(Consumer &consumer)
 
         it = consumer.m_toSync.erase(it);
     }
+    
+    if (!replayDone && WarmStart::isWarmStart() && m_pendingReplayIntfList.empty() )
+    {
+        replayDone = true;
+        WarmStart::setWarmStartState("intfmgrd", WarmStart::REPLAYED);
+        // There is no operation to be performed for intfmgr reconcillation
+        // Hence mark it reconciled right away
+        WarmStart::setWarmStartState("intfmgrd", WarmStart::RECONCILED);
+    }
 }

cfgmgr/intfmgr.h

@@ -19,10 +19,12 @@ class IntfMgr : public Orch
 
 private:
     ProducerStateTable m_appIntfTableProducer;
+    Table m_cfgIntfTable, m_cfgVlanIntfTable, m_cfgLagIntfTable, m_cfgLoopbackIntfTable;
     Table m_statePortTable, m_stateLagTable, m_stateVlanTable, m_stateVrfTable, m_stateIntfTable;
 
     std::set<std::string> m_subIntfList;
     std::set<std::string> m_loopbackIntfList;
+    std::set<std::string> m_pendingReplayIntfList;
 
     void setIntfIp(const std::string &alias, const std::string &opCmd, const IpPrefix &ipPrefix);
     void setIntfVrf(const std::string &alias, const std::string &vrfName);
@@ -36,6 +38,7 @@ class IntfMgr : public Orch
     bool isIntfCreated(const std::string &alias);
     bool isIntfChangeVrf(const std::string &alias, const std::string &vrfName);
     int getIntfIpCount(const std::string &alias);
+    void buildIntfReplayList(void);
 
     void addLoopbackIntf(const std::string &alias);
     void delLoopbackIntf(const std::string &alias);

cfgmgr/vrfmgr.cpp

@@ -102,6 +102,11 @@ VrfMgr::VrfMgr(DBConnector *cfgDb, DBConnector *appDb, DBConnector *stateDb, con
             << IP_CMD << " -6 rule add pref " << TABLE_LOCAL_PREF << " table local && " << IP_CMD << " -6 rule del pref 0";
         EXEC_WITH_ERROR_THROW(cmd.str(), res);
     }
+
+    if (!WarmStart::isWarmStart())
+    {
+        WarmStart::setWarmStartState("vrfmgrd", WarmStart::WSDISABLED);
+    }
 }
 
 uint32_t VrfMgr::getFreeTable(void)

cfgmgr/vrfmgrd.cpp

@@ -35,6 +35,7 @@ mutex gDbMutex;
 int main(int argc, char **argv)
 {
     Logger::linkToDbNative("vrfmgrd");
+    bool isWarmStart = false;
     SWSS_LOG_ENTER();
 
     SWSS_LOG_NOTICE("--- Starting vrfmgrd ---");
@@ -56,6 +57,8 @@ int main(int argc, char **argv)
 
         VrfMgr vrfmgr(&cfgDb, &appDb, &stateDb, cfg_vrf_tables);
 
+        isWarmStart = WarmStart::isWarmStart();
+
         // TODO: add tables in stateDB which interface depends on to monitor list
         std::vector<Orch *> cfgOrchList = {&vrfmgr};
 
@@ -69,6 +72,7 @@ int main(int argc, char **argv)
         while (true)
         {
             Selectable *sel;
+            static bool firstReadTimeout = true;
             int ret;
 
             ret = s.select(&sel, SELECT_TIMEOUT);
@@ -80,6 +84,14 @@ int main(int argc, char **argv)
             if (ret == Select::TIMEOUT)
             {
                 vrfmgr.doTask();
+                if (isWarmStart && firstReadTimeout)
+                {
+                    firstReadTimeout = false;
+                    WarmStart::setWarmStartState("vrfmgrd", WarmStart::REPLAYED);
+                    // There is no operation to be performed for vrfmgrd reconcillation
+                    // Hence mark it reconciled right away
+                    WarmStart::setWarmStartState("vrfmgrd", WarmStart::RECONCILED);
+                }
                 continue;
             }
 

tests/test_warm_reboot.py

@@ -46,7 +46,7 @@ def swss_check_RestoreCount(dvs, state_db, restore_count):
             if fv[0] == "restore_count":
                 assert int(fv[1]) == restore_count[key] + 1
             elif fv[0] == "state":
-                assert fv[1] == "reconciled"
+                assert fv[1] == "reconciled" or fv[1] == "disabled"
 
 def check_port_oper_status(appl_db, port_name, state):
     portTbl = swsscommon.Table(appl_db, swsscommon.APP_PORT_TABLE_NAME)
@@ -76,7 +76,7 @@ def swss_app_check_RestoreCount_single(state_db, restore_count, name):
             if fv[0] == "restore_count":
                 assert int(fv[1]) == restore_count[key] + 1
             elif fv[0] == "state":
-                assert fv[1] == "reconciled"
+                assert fv[1] == "reconciled"  or fv[1] == "disabled"
 
 def swss_app_check_warmstart_state(state_db, name, state):
     warmtbl = swsscommon.Table(state_db, swsscommon.STATE_WARM_RESTART_TABLE_NAME)