Fix token leak to other cluster during forced node add

CHANGELOG.md

@@ -23,6 +23,8 @@
   alphabetically ([rhbz#2461143]) ([RHEL-176475])
 - Location constraints with rules can be removed using web UI (broken since
   pcs-0.12.0b1) ([RHEL-183182])
+- Adding a node to a cluster now removes all cluster configuration files from
+  the new node before adding it.
 
 ### Deprecated
 - Commands `pcs pcsd certkey` and `pcs pcsd sync-certificates` ([RHEL-149608])

pcs/lib/commands/cluster/setup_node.py

@@ -19,6 +19,7 @@
     start_cluster,
 )
 from pcs.lib.commands.cluster.utils import ensure_live_env, verify_corosync_conf
+from pcs.lib.communication import cluster
 from pcs.lib.communication.corosync import (
     CheckCorosyncOffline,
     DistributeCorosyncConf,
@@ -357,7 +358,14 @@ def add_nodes(  # noqa: PLR0912, PLR0915
     # Validation done. If errors occurred, an exception has been raised and we
     # don't get below this line.
 
-    # First set up everything else than corosync. Once the new nodes are present
+    # First, destroy cluster on new nodes. This is needed to make sure that
+    # new nodes are not part of another cluster and that there are no cluster
+    # configs left there which would interfere with the current cluster.
+    com_cmd = cluster.Destroy(env.report_processor)
+    com_cmd.set_targets(new_nodes_target_list)
+    run_and_raise(env.get_node_communicator(), com_cmd)
+
+    # Set up everything else than corosync. Once the new nodes are present
     # in corosync.conf, they're considered part of a cluster and the node add
     # command cannot be run again. So we need to minimize the amount of actions
     # (and therefore possible failures) after adding the nodes to corosync.