Fix confluentinc#339 - make available() a no-op (confluentinc#340) (c…

…onfluentinc#343)

* dont use copy

* dont use continue

* also log scheduling errors

* small fixes

* make available() a noop in remoteaccount.py

* bump version to 0.10.1 + changelog

* style

docs/changelog.rst

@@ -4,6 +4,15 @@
 Changelog
 ====
 
+0.10.1
+======
+- Disable health checks for nodes, effectively disabling `#325 <https://github.com/confluentinc/ducktape/pull/325>`_. See github issue for details - `#339 <https://github.com/confluentinc/ducktape/issues/339>`_
+
+0.10.0
+======
+- **DO NOT USE**, this release has a nasty bug - `#339 <https://github.com/confluentinc/ducktape/issues/339>`_
+- Do not schedule tests on unresponsive nodes - `#325 <https://github.com/confluentinc/ducktape/pull/325>`_
+
 0.9.2
 =====
 - Service release, no ducktape changes, simply fixed readthedocs configs.

ducktape/__init__.py

@@ -1 +1 @@
-__version__ = '0.9.2'
+__version__ = '0.10.1'

ducktape/cluster/json.py

@@ -116,7 +116,7 @@ def do_alloc(self, cluster_spec):
         try:
             good_nodes, bad_nodes = self._available_accounts.remove_spec(cluster_spec)
         except InsufficientHealthyNodesError as e:
-            self._bad_accounts.add_nodes(e)
+            self._bad_accounts.add_nodes(e.bad_nodes)
             raise e
 
         # even in case of no exceptions, we can still run into bad nodes, so let's track them

ducktape/cluster/remoteaccount.py

@@ -279,14 +279,16 @@ def _can_ping_url(self, url, headers):
             return False
 
     def available(self):
-        try:
-            self.ssh_client
-        except Exception:
-            return False
-        else:
-            return True
-        finally:
-            self.close()
+        # TODO: https://github.com/confluentinc/ducktape/issues/339
+        # try:
+        #     self.ssh_client
+        # except Exception:
+        #     return False
+        # else:
+        #     return True
+        # finally:
+        #     self.close()
+        return True
 
     @check_ssh
     def ssh(self, cmd, allow_fail=False):

ducktape/tests/runner.py

@@ -217,12 +217,16 @@ def run_all_tests(self):
                         # free up healthy nodes.
                         # However, if some nodes failed, cluster size changed too, so we need to check if
                         # there are any tests that can no longer be scheduled.
+                        self._log(
+                            logging.INFO,
+                            f"Couldn't schedule test context {next_test_context} but we'll keep trying",
+                            exc_info=True
+                        )
                         self._check_unschedulable()
-                        continue
-
-                    # only remove the test from the scheduler when we have successfully allocated a subcluster for it
-                    self.scheduler.remove(next_test_context)
-                    self._run_single_test(next_test_context)
+                    else:
+                        # only remove the test from the scheduler once we've successfully allocated a subcluster for it
+                        self.scheduler.remove(next_test_context)
+                        self._run_single_test(next_test_context)
 
                 if self._expect_client_requests:
                     try:

systests/cluster/test_debug.py

@@ -5,7 +5,7 @@
 """
 import time
 
-from ducktape.mark import matrix, parametrize
+from ducktape.mark import matrix, parametrize, ignore
 from ducktape.mark.resource import cluster
 from ducktape.tests.test import Test
 from systests.cluster.test_remote_account import GenericService
@@ -71,6 +71,11 @@ def another_two_node_test(self):
         self.service = GenericService(self.test_context, 2)
         assert True
 
+    @ignore
+    @cluster(num_nodes=2)
+    def a_two_node_ignored_test(self):
+        assert False
+
     @cluster(num_nodes=2)
     def yet_another_two_node_test(self):
         self.service = GenericService(self.test_context, 2)