Redistribute master key on downgrade

When downgrading from 2.13 to 2.12, the SSH key pair of
each node needs to be replaced by the master's SSH key
pair. If that is not done, any node added after the
downgrade will not be reachable if the master is failed
over to one of the original non-master nodes.

This patch fixes Issue 1008. However, what is not cleaned
up is the nodes' authorized_keys file. This will need
significantly more effort, but this patch at least fixes
the user-impacting issue.

Signed-off-by: Helga Velroyen <helgav@google.com>
Reviewed-by: Klaus Aehlig <aehlig@google.com>

diff --git a/lib/client/gnt_cluster.py b/lib/client/gnt_cluster.py
index 6165dce..7304aab 100644 (file)
--- a/lib/client/gnt_cluster.py
+++ b/lib/client/gnt_cluster.py
@@ -2181,6 +2181,42 @@ def _VersionSpecificDowngrade():
   """
   ToStdout("Performing version-specific downgrade tasks.")
 
+  # Determine if this cluster is set up with SSH handling
+  # (aka not using --no-ssh-init), check if the public
+  # keyfile exists.
+  update_keys = os.path.exists(pathutils.SSH_PUB_KEYS)
+
+  if not update_keys:
+    return True
+
+  ToStdout("Replace nodes' SSH keys with the master's keys.")
+  (_, root_keyfiles) = \
+    ssh.GetAllUserFiles(constants.SSH_LOGIN_USER, mkdir=False, dircheck=False)
+
+  dsa_root_keyfiles = dict((kind, value) for (kind, value)
+                           in root_keyfiles.items()
+                           if kind == constants.SSHK_DSA)
+  master_private_keyfile, master_public_keyfile = \
+      dsa_root_keyfiles[constants.SSHK_DSA]
+
+  nodes = ssconf.SimpleStore().GetOnlineNodeList()
+  master_node = ssconf.SimpleStore().GetMasterNode()
+  cluster_name = ssconf.SimpleStore().GetClusterName()
+
+  # If master node is in 'nodes', remove it
+  if master_node in nodes:
+    nodes.remove(master_node)
+
+  srun = ssh.SshRunner(cluster_name=cluster_name)
+  for name in nodes:
+    for key_file in [master_private_keyfile, master_public_keyfile]:
+      command = utils.text.ShellQuoteArgs([
+          "scp", key_file, "%s:%s" % (name, key_file)])
+      result = srun.Run(master_node, constants.SSH_LOGIN_USER, command)
+      if result.exit_code != 0:
+        ToStderr("Overiding SSH key '%s' of node '%s' failed. You might"
+                 " want to clean up manually." % (key_file, name))
+
   return True