From d67bc06645aed1529493885a0699d99a0e129b86 Mon Sep 17 00:00:00 2001
From: Greg Goodrich <ggoodrich@ippathways.com>
Date: Mon, 24 Feb 2020 09:38:55 -0600
Subject: [PATCH] Incorporate VR OOB start checks to properly HA the VR

---
 .../VirtualNetworkApplianceManagerImpl.java   | 45 ++++++++++++++++++-
 1 file changed, 44 insertions(+), 1 deletion(-)

diff --git a/server/src/main/java/com/cloud/network/router/VirtualNetworkApplianceManagerImpl.java b/server/src/main/java/com/cloud/network/router/VirtualNetworkApplianceManagerImpl.java
index 7ff911393f1a..72f08ec4001b 100644
--- a/server/src/main/java/com/cloud/network/router/VirtualNetworkApplianceManagerImpl.java
+++ b/server/src/main/java/com/cloud/network/router/VirtualNetworkApplianceManagerImpl.java
@@ -384,6 +384,7 @@ public class VirtualNetworkApplianceManagerImpl extends ManagerBase implements V
     private ScheduledExecutorService _checkExecutor;
     private ScheduledExecutorService _networkStatsUpdateExecutor;
     private ExecutorService _rvrStatusUpdateExecutor;
+    private ExecutorService _routerOobStartExecutor;
 
     private BlockingQueue<Long> _vrUpdateQueue;
 
@@ -544,6 +545,7 @@ public boolean configure(final String name, final Map<String, Object> params) th
         _executor = Executors.newScheduledThreadPool(1, new NamedThreadFactory("RouterMonitor"));
         _checkExecutor = Executors.newScheduledThreadPool(1, new NamedThreadFactory("RouterStatusMonitor"));
         _networkStatsUpdateExecutor = Executors.newScheduledThreadPool(1, new NamedThreadFactory("NetworkStatsUpdater"));
+        _routerOobStartExecutor = Executors.newCachedThreadPool(new NamedThreadFactory("CheckRouterAfterOobStartExecutor"));
 
         VirtualMachine.State.getStateMachine().registerListener(this);
 
@@ -3279,7 +3281,12 @@ public boolean postStateTransitionEvent(final StateMachine2.Transition<VirtualMa
                 event == VirtualMachine.Event.FollowAgentPowerOnReport &&
                 newState == VirtualMachine.State.Running &&
                 isOutOfBandMigrated(opaque)) {
-            s_logger.debug("Virtual router " + vo.getInstanceName() + " is powered-on out-of-band");
+            // Since vRouter appears to be powered-on OOB, make sure we can talk to router
+            // If we can't talk to it, we need to reboot it to get it managed correctly
+            // This is needed for example when a host agent goes down and comes back up,
+            // we would have done a failed HA event on the router and end up having our controlIP out-of-sync
+            s_logger.info("Virtual router " + vo.getInstanceName() + " is powered-on out-of-band, checking if can send CheckRouterCommand to router");
+            _routerOobStartExecutor.execute(new CheckRouterAfterOobStart(vo.getId()));
         }
 
         return true;
@@ -3309,6 +3316,42 @@ private boolean isOutOfBandMigrated(final Object opaque) {
         return false;
     }
 
+    protected class CheckRouterAfterOobStart extends ManagedContextRunnable {
+
+        long _routerId;
+
+        public CheckRouterAfterOobStart(final long routerId) {
+            _routerId = routerId;
+        }
+
+        @Override
+        protected void runInContext() {
+            // Since vRouter appears to be powered-on OOB, make sure we can talk to router
+            // If we can't talk to it, we need to reboot it to get it managed correctly
+            // This is needed for example when a host agent goes down and comes back up,
+            // we would have done a failed HA event on the router and end up having our controlIP out-of-sync
+            final DomainRouterVO router = _routerDao.findById(_routerId);
+            final CheckRouterCommand command = new CheckRouterCommand();
+            final String routerDesc = router.getInstanceName() + " (ID:" + _routerId + ")";
+
+            command.setAccessDetail(NetworkElementCommand.ROUTER_IP, _routerControlHelper.getRouterControlIp(_routerId));
+            command.setAccessDetail(NetworkElementCommand.ROUTER_NAME, router.getInstanceName());
+            command.setWait(30);
+            final Answer answer = _agentMgr.easySend(router.getHostId(), command);
+            if (answer instanceof CheckRouterAnswer && answer.getResult()) {
+                s_logger.info("Successfully able to send CheckRouterCommand to " + routerDesc + " after out-of-band power-on");
+            }
+            else {
+                s_logger.warn("Unable to send CheckRouterCommand to " + routerDesc + ", rebooting router ");
+                try {
+                    rebootRouter(_routerId, true, false);
+                } catch (final Exception e) {
+                    s_logger.error("Error while rebooting router " + routerDesc, e);
+                }
+            }
+        }
+    }
+
     protected boolean aggregationExecution(final AggregationControlCommand.Action action, final Network network, final List<DomainRouterVO> routers)
             throws AgentUnavailableException, ResourceUnavailableException {