Propagate Boto exception instead of overwriting. (#15564)

python/ray/autoscaler/_private/aws/config.py

@@ -337,11 +337,7 @@ def _configure_key_pair(config):
             "No matching local key file for any of the key pairs in this "
             "account with ids from 0..{}. "
             "Consider deleting some unused keys pairs from your account.",
-            key_name)  # todo: err msg
+            key_name)
-        raise ValueError(
-            "No matching local key file for any of the key pairs in this "
-            "account with ids from 0..{}. ".format(key_name) +
-            "Consider deleting some unused keys pairs from your account.")
 
     cli_logger.doassert(
         os.path.exists(key_path), "Private key file " + cf.bold("{}") +
@@ -391,13 +387,8 @@ def _configure_subnet(config):
             "and trying this again.\n"
             "Note that the subnet must map public IPs "
             "on instance launch unless you set `use_internal_ips: true` in "
-            "the `provider` config.")  # todo: err msg
+            "the `provider` config.")
-        raise Exception(
+
-            "No usable subnets found, try manually creating an instance in "
-            "your specified region to populate the list of subnets "
-            "and trying this again. Note that the subnet must map public IPs "
-            "on instance launch unless you set 'use_internal_ips': True in "
-            "the 'provider' config.")
     if "availability_zone" in config["provider"]:
         azs = config["provider"]["availability_zone"].split(",")
         subnets = [s for s in subnets if s.availability_zone in azs]
@@ -407,13 +398,7 @@ def _configure_subnet(config):
                 "Choose a different availability zone or try "
                 "manually creating an instance in your specified region "
                 "to populate the list of subnets and trying this again.",
-                config["provider"]["availability_zone"])  # todo: err msg
+                config["provider"]["availability_zone"])
-            raise Exception(
-                "No usable subnets matching availability zone {} "
-                "found. Choose a different availability zone or try "
-                "manually creating an instance in your specified region "
-                "to populate the list of subnets and trying this again.".
-                format(config["provider"]["availability_zone"]))
 
     # Use subnets in only one VPC, so that _configure_security_groups only
     # needs to create a security group in this one VPC. Otherwise, we'd need

python/ray/autoscaler/_private/aws/node_provider.py

@@ -399,12 +399,12 @@ class AWSNodeProvider(NodeProvider):
                 break
             except botocore.exceptions.ClientError as exc:
                 if attempt == BOTO_CREATE_MAX_RETRIES:
-                    # todo: err msg
                     cli_logger.abort(
-                        "Failed to launch instances. Max attempts exceeded.")
+                        "Failed to launch instances. Max attempts exceeded.",
-                    raise exc
+                        exc=exc,
+                    )
                 else:
-                    cli_logger.print(
+                    cli_logger.warning(
                         "create_instances: Attempt failed with {}, retrying.",
                         exc)
         return created_nodes_dict

python/ray/autoscaler/_private/cli_logger.py

@@ -610,7 +610,10 @@ class _CliLogger():
         exc_cls = click.ClickException
         if self.pretty:
             exc_cls = SilentClickException
-        raise exc_cls("Exiting due to cli_logger.abort()")
+
+        if msg is None:
+            msg = "Exiting due to cli_logger.abort()"
+        raise exc_cls(msg)
 
     def doassert(self, val: bool, msg: str, *args: Any, **kwargs: Any):
         """Handle assertion without throwing a scary exception.

python/ray/autoscaler/_private/commands.py

@@ -190,7 +190,6 @@ def create_or_update_cluster(
         cli_logger.abort(
             "Provided cluster configuration file ({}) does not exist",
             cf.bold(config_file))
-        raise
     except yaml.parser.ParserError as e:
         handle_yaml_error(e)
         raise
@@ -211,8 +210,6 @@ def create_or_update_cluster(
                 k for k in _NODE_PROVIDERS.keys()
                 if _NODE_PROVIDERS[k] is not None
             ]))
-        raise NotImplementedError("Unsupported provider {}".format(
-            config["provider"]))
 
     printed_overrides = False
 
@@ -644,9 +641,8 @@ def get_or_create_head_node(config: Dict[str, Any],
             with cli_logger.group("Fetching the new head node"):
                 while True:
                     if time.time() - start > 50:
-                        cli_logger.abort(
+                        cli_logger.abort("Head node fetch timed out. "
-                            "Head node fetch timed out.")  # todo: msg
+                                         "Failed to create head node.")
-                        raise RuntimeError("Failed to create head node.")
                     nodes = provider.non_terminated_nodes(head_node_tags)
                     if len(nodes) == 1:
                         head_node = nodes[0]

python/ray/autoscaler/_private/updater.py

@@ -120,7 +120,6 @@ class NodeUpdater:
                    "Either do not pass `--redirect-command-output` "
                    "or also pass in `--use-normal-shells`.")
             cli_logger.abort(msg)
-            raise click.ClickException(msg)
 
         try:
             with LogTimer(self.log_prefix +