scylla_node: add missing liveness check

When starting a node with "wait_other_notice", the existing code waited until all other nodes think the new node is ready. However, it's safer to also check that the new node thinks all the other nodes are ready. It should usually be the case (the new node gets information on all other nodes in the first gossip), but it doesn't strictly have to be true so better ensure that it is. Signed-off-by: Nadav Har'El <[email protected]>
fruch · Jul 19, 2023 · 65872f6 · 65872f6
1 parent 757c036
commit 65872f6
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/ccmlib/scylla_node.py b/ccmlib/scylla_node.py
@@ -311,6 +311,7 @@ def _start_scylla(self, args, marks, update_pid, wait_other_notice,
             for node, _ in marks:
                 t = timeout if timeout is not None else 120 if self.cluster.scylla_mode != 'debug' else 360
                 node.watch_rest_for_alive(self, timeout=t)
+                self.watch_rest_for_alive(node, timeout=t)
 
         if wait_for_binary_proto:
             t = timeout * 4 if timeout is not None else 420 if self.cluster.scylla_mode != 'debug' else 900