Distinguish different CLA instances during restart

There may be a situation where there are multiple PeerDisappeared-messages for the same CLA waiting to be processed.
This might cause the manager to restart the same CLA multiple times in
quick succession.
To prevent this, we can check if the pointer to the CLA from the
PeerDisappeared-message points to the same instance as the one currently
stored in the manager's `convs` map.

pkg/cla/manager.go

@@ -1,5 +1,7 @@
 // SPDX-FileCopyrightText: 2019, 2020 Alvar Penning
-// SPDX-FileCopyrightText: 2020 Markus Sommer
+// SPDX-FileCopyrightText: 2020, 2021 Markus Sommer
+// SPDX-FileCopyrightText: 2021 Artur Sterz
+// SPDX-FileCopyrightText: 2021 Jonas Höchst
 //
 // SPDX-License-Identifier: GPL-3.0-or-later
 
@@ -266,7 +268,18 @@ func (manager *Manager) unregisterConvergence(conv Convergence) {
 		return
 	}
 
-	convElem.(*convergenceElem).deactivate(manager.queueTtl)
+	element := convElem.(*convergenceElem)
+
+	if element.conv != conv {
+		log.WithFields(log.Fields{
+			"cla":     conv,
+			"address": conv.Address(),
+		}).Error("CLA not unregistered, different instance.")
+
+		return
+	}
+
+	element.deactivate(manager.queueTtl)
 	manager.convs.Delete(conv.Address())
 }