Support warm-restart for neighsyncd (#599)

* Support warm-restart for neighsyncd
- No function changes if warm-restart is disabled.
In case warm-restart is enabled, below is done in the process to make sure the ARP/NDP is up-to-date:
- ARP/NDP will be restored from App-DB marked with stale.
- new/changed ARP/DNP during the warm-restart marked as new
- deleted ARP/NDP during the warm-restart marked as delete
- same ARP/NDP entries marked as same
- reconciliation is done after timer-out of a timer
- delta (including stale, new, delete) during the warm-restart would be pushed to App-DB and then to HW.

Add vs test cases to cover:
- No changes during the warm-restart
- ADD/DELETE/CHANGE/SAME neighbors during the warm-restart
- verify the timer changes taking effect for neighsyncd

* Add ps-table clear before reading table, address a few comments

* Added a few improvements

* Addressed a few comments

* Moved the class to a common place

* Addressed the review feedback

* Removed some unused code and fixed a typo

neighsyncd/Makefile.am

@@ -1,4 +1,4 @@
-INCLUDES = -I $(top_srcdir)
+INCLUDES = -I $(top_srcdir) -I $(top_srcdir)/warmrestart
 
 bin_PROGRAMS = neighsyncd
 
@@ -8,7 +8,7 @@ else
 DBGFLAGS = -g
 endif
 
-neighsyncd_SOURCES = neighsyncd.cpp neighsync.cpp
+neighsyncd_SOURCES = neighsyncd.cpp neighsync.cpp $(top_srcdir)/warmrestart/warm_restart.cpp $(top_srcdir)/warmrestart/warmRestartAssist.cpp
 
 neighsyncd_CFLAGS = $(DBGFLAGS) $(AM_CFLAGS) $(CFLAGS_COMMON)
 neighsyncd_CPPFLAGS = $(DBGFLAGS) $(AM_CFLAGS) $(CFLAGS_COMMON)

neighsyncd/neighsync.cpp

@@ -11,12 +11,14 @@
 #include "linkcache.h"
 
 #include "neighsync.h"
+#include "warm_restart.h"
 
 using namespace std;
 using namespace swss;
 
-NeighSync::NeighSync(DBConnector *db) :
-    m_neighTable(db, APP_NEIGH_TABLE_NAME)
+NeighSync::NeighSync(RedisPipeline *pipelineAppDB) :
+    m_neighTable(pipelineAppDB, APP_NEIGH_TABLE_NAME),
+    m_AppRestartAssist(pipelineAppDB, "neighsyncd", "swss", &m_neighTable, DEFAULT_NEIGHSYNC_WARMSTART_TIMER)
 {
 }
 
@@ -52,18 +54,32 @@ void NeighSync::onMsg(int nlmsg_type, struct nl_object *obj)
     key+= ipStr;
 
     int state = rtnl_neigh_get_state(neigh);
+    bool delete_key = false;
     if ((nlmsg_type == RTM_DELNEIGH) || (state == NUD_INCOMPLETE) ||
         (state == NUD_FAILED))
     {
-        m_neighTable.del(key);
-        return;
+	    delete_key = true;
     }
 
     nl_addr2str(rtnl_neigh_get_lladdr(neigh), macStr, MAX_ADDR_SIZE);
+
     std::vector<FieldValueTuple> fvVector;
     FieldValueTuple f("family", family);
     FieldValueTuple nh("neigh", macStr);
     fvVector.push_back(nh);
     fvVector.push_back(f);
-    m_neighTable.set(key, fvVector);
+
+    if (m_AppRestartAssist.isWarmStartInProgress())
+    {
+        m_AppRestartAssist.insertToMap(key, fvVector, delete_key);
+    }
+    else
+    {
+        if (delete_key == true)
+        {
+            m_neighTable.del(key);
+            return;
+        }
+        m_neighTable.set(key, fvVector);
+    }
 }

neighsyncd/neighsync.h

@@ -4,6 +4,9 @@
 #include "dbconnector.h"
 #include "producerstatetable.h"
 #include "netmsg.h"
+#include "warmRestartAssist.h"
+
+#define DEFAULT_NEIGHSYNC_WARMSTART_TIMER 5
 
 namespace swss {
 
@@ -12,12 +15,18 @@ class NeighSync : public NetMsg
 public:
     enum { MAX_ADDR_SIZE = 64 };
 
-    NeighSync(DBConnector *db);
+    NeighSync(RedisPipeline *pipelineAppDB);
 
     virtual void onMsg(int nlmsg_type, struct nl_object *obj);
 
+    AppRestartAssist *getRestartAssist()
+    {
+        return &m_AppRestartAssist;
+    }
+
 private:
     ProducerStateTable m_neighTable;
+    AppRestartAssist m_AppRestartAssist;
 };
 
 }

neighsyncd/neighsyncd.cpp

@@ -11,8 +11,11 @@ using namespace swss;
 int main(int argc, char **argv)
 {
     Logger::linkToDbNative("neighsyncd");
-    DBConnector db(APPL_DB, DBConnector::DEFAULT_UNIXSOCKET, 0);
-    NeighSync sync(&db);
+
+    DBConnector appDb(APPL_DB, DBConnector::DEFAULT_UNIXSOCKET, 0);
+    RedisPipeline pipelineAppDB(&appDb);
+
+    NeighSync sync(&pipelineAppDB);
 
     NetDispatcher::getInstance().registerMessageHandler(RTM_NEWNEIGH, &sync);
     NetDispatcher::getInstance().registerMessageHandler(RTM_DELNEIGH, &sync);
@@ -29,10 +32,23 @@ int main(int argc, char **argv)
             netlink.dumpRequest(RTM_GETNEIGH);
 
             s.addSelectable(&netlink);
+            if (sync.getRestartAssist()->isWarmStartInProgress())
+            {
+                sync.getRestartAssist()->readTableToMap();
+                sync.getRestartAssist()->startReconcileTimer(s);
+            }
             while (true)
             {
                 Selectable *temps;
                 s.select(&temps);
+                if (sync.getRestartAssist()->isWarmStartInProgress())
+                {
+                    if (sync.getRestartAssist()->checkReconcileTimer(temps))
+                    {
+                        sync.getRestartAssist()->stopReconcileTimer(s);
+                        sync.getRestartAssist()->reconcile();
+                    }
+                }
             }
         }
         catch (const std::exception& e)