[WR]: Add reconciliation logic for teamsyncd (sonic-net#725)

pavel-shirshov · qiluo-msft · commit 025268aafffd · 2018-12-22T23:09:21.000-08:00
* Pospone QueueMap initialization until activation of counters
* Generate queue maps only for front panel ports
* Initial commit for teamsyncd reconc logic
* Add log messages
* PR refactoring
* Make pending timeout configurable
* if space thing
* Update schema
diff --git a/doc/swss-schema.md b/doc/swss-schema.md
@@ -714,6 +714,13 @@ Stores information for physical switch ports managed by the switch chip. Ports t
                                             ; the elected routing-stack.
                                             ; Supported range: 1-3600.
 
+    teamsyncd_timer     = 1*4DIGIT          ; teamsyncd_timer holds the time interval utilized by teamsyncd during warm-restart episodes.
+                                            ; The timer is started when teamsyncd starts. During the timer interval teamsyncd 
+                                            ; will preserver all LAG interface changes, but it will not apply them. The changes 
+                                            ; will only be applied when the timer expired. During the changes application the stale
+                                            ; LAG entries will be removed, the new LAG entries will be created.
+                                            ; Supported range: 1-9999. 0 is invalid
+
 
 ### VXLAN\_TUNNEL
 Stores vxlan tunnels configuration
diff --git a/teamsyncd/teamsync.cpp b/teamsyncd/teamsync.cpp
@@ -4,13 +4,16 @@
 #include <sys/socket.h>
 #include <linux/if.h>
 #include <netlink/route/link.h>
+#include <chrono>
 #include "logger.h"
 #include "netmsg.h"
 #include "dbconnector.h"
 #include "producerstatetable.h"
+#include "warm_restart.h"
 #include "teamsync.h"
 
 using namespace std;
+using namespace std::chrono;
 using namespace swss;
 
 /* Taken from drivers/net/team/team.c */
@@ -22,6 +25,34 @@ TeamSync::TeamSync(DBConnector *db, DBConnector *stateDb, Select *select) :
     m_lagMemberTable(db, APP_LAG_MEMBER_TABLE_NAME),
     m_stateLagTable(stateDb, STATE_LAG_TABLE_NAME)
 {
+    WarmStart::initialize("teamsyncd", "teamd");
+    WarmStart::checkWarmStart("teamsyncd", "teamd");
+    m_warmstart = WarmStart::isWarmStart();
+
+    if (m_warmstart)
+    {
+        m_start_time = steady_clock::now();
+        auto warmRestartIval = WarmStart::getWarmStartTimer("teamsyncd", "teamd");
+        m_pending_timeout = warmRestartIval ? warmRestartIval : DEFAULT_WR_PENDING_TIMEOUT;
+        m_lagTable.create_temp_view();
+        m_lagMemberTable.create_temp_view();
+        SWSS_LOG_NOTICE("Starting in warmstart mode");
+    }
+}
+
+void TeamSync::periodic()
+{
+    if (m_warmstart)
+    {
+        auto diff = duration_cast<seconds>(steady_clock::now() - m_start_time);
+        if(diff.count() > m_pending_timeout)
+        {
+            applyState();
+            m_warmstart = false; // apply state just once
+        }
+    }
+
+    doSelectableTask();
 }
 
 void TeamSync::doSelectableTask()
@@ -44,6 +75,23 @@ void TeamSync::doSelectableTask()
     m_selectablesToRemove.clear();
 }
 
+void TeamSync::applyState()
+{
+    SWSS_LOG_NOTICE("Applying state");
+
+    m_lagTable.apply_temp_view();
+    m_lagMemberTable.apply_temp_view();
+
+    for(auto &it: m_stateLagTablePreserved)
+    {
+        const auto &lagName  = it.first;
+        const auto &fvVector = it.second;
+        m_stateLagTable.set(lagName, fvVector);
+    }
+
+    m_stateLagTablePreserved.clear();
+}
+
 void TeamSync::onMsg(int nlmsg_type, struct nl_object *obj)
 {
     struct rtnl_link *link = (struct rtnl_link *)obj;
@@ -90,7 +138,14 @@ void TeamSync::addLag(const string &lagName, int ifindex, bool admin_state,
     fvVector.clear();
     FieldValueTuple s("state", "ok");
     fvVector.push_back(s);
-    m_stateLagTable.set(lagName, fvVector);
+    if (m_warmstart)
+    {
+        m_stateLagTablePreserved[lagName] = fvVector;
+    }
+    else
+    {
+        m_stateLagTable.set(lagName, fvVector);
+    }
 
     /* Create the team instance */
     auto sync = make_shared<TeamPortSync>(lagName, ifindex, &m_lagMemberTable);
@@ -116,7 +171,15 @@ void TeamSync::removeLag(const string &lagName)
     if (m_teamSelectables.find(lagName) == m_teamSelectables.end())
         return;
 
-    m_stateLagTable.del(lagName);
+    if (m_warmstart)
+    {
+        m_stateLagTablePreserved.erase(lagName);
+    }
+    else
+    {
+        m_stateLagTable.del(lagName);
+    }
+
     m_selectablesToRemove.insert(lagName);
 }
 
diff --git a/teamsyncd/teamsync.h b/teamsyncd/teamsync.h
@@ -11,19 +11,23 @@
 #include "netmsg.h"
 #include <team.h>
 
+// seconds
+const uint32_t DEFAULT_WR_PENDING_TIMEOUT = 70;
+
+using namespace std::chrono;
+
 namespace swss {
 
 class TeamSync : public NetMsg
 {
 public:
     TeamSync(DBConnector *db, DBConnector *stateDb, Select *select);
 
+    void periodic();
+
     /* Listen to RTM_NEWLINK, RTM_DELLINK to track team devices */
     virtual void onMsg(int nlmsg_type, struct nl_object *obj);
 
-    /* Handle all selectables add/removal events */
-    void doSelectableTask();
-
     class TeamPortSync : public Selectable
     {
     public:
@@ -54,12 +58,23 @@ class TeamSync : public NetMsg
                 bool oper_state);
     void removeLag(const std::string &lagName);
 
+    /* valid only in WR mode */
+    void applyState();
+
+    /* Handle all selectables add/removal events */
+    void doSelectableTask();
+
 private:
     Select *m_select;
     ProducerStateTable m_lagTable;
     ProducerStateTable m_lagMemberTable;
     Table m_stateLagTable;
 
+    bool m_warmstart;
+    std::unordered_map<std::string, std::vector<FieldValueTuple>> m_stateLagTablePreserved;
+    steady_clock::time_point m_start_time;
+    uint32_t m_pending_timeout;
+
     /* Store selectables needed to be updated in doSelectableTask function */
     std::set<std::string> m_selectablesToAdd;
     std::set<std::string> m_selectablesToRemove;
diff --git a/teamsyncd/teamsyncd.cpp b/teamsyncd/teamsyncd.cpp
@@ -34,9 +34,8 @@ int main(int argc, char **argv)
             while (true)
             {
                 Selectable *temps;
-                s.select(&temps);
-
-                sync.doSelectableTask();
+                s.select(&temps, 1000); // block for a second
+                sync.periodic();
             }
         }
         catch (const std::exception& e)

Original file line number	Diff line number	Diff line change
`@@ -34,9 +34,8 @@ int main(int argc, char **argv)`
`34`	`34`	`while (true)`
`35`	`35`	`{`
`36`	`36`	`Selectable *temps;`
`37`		`- s.select(&temps);`
`38`		`-`
`39`		`- sync.doSelectableTask();`
	`37`	`+ s.select(&temps, 1000); // block for a second`
	`38`	`+ sync.periodic();`
`40`	`39`	`}`
`41`	`40`	`}`
`42`	`41`	`catch (const std::exception& e)`