replicaset: more informative error at bootstrap

CHANGELOG.md

@@ -18,6 +18,8 @@ and this project adheres to [Semantic Versioning](http://semver.org/spec/v2.0.0.
 - `tt cluster failover switch`: minor change in output that displays corresponding
   `switch-status` command with quoted URI argument so it could be copy-pasted for
   subsequent launch as-is.
+- `tt rs vshard bootstrap`: make more informative error message when sharding roles
+  are not configured (for example when launched against non-vshard cluster).
 
 ### Fixed
 

cli/replicaset/cconfig.go

@@ -26,7 +26,8 @@ var (
 	//go:embed lua/cconfig/bootstrap_vshard_body.lua
 	cconfigBootstrapVShardBody string
 
-	cconfigGetShardingRolesBody = "return require('config'):get().sharding.roles"
+	//go:embed lua/cconfig/get_sharding_roles_body.lua
+	cconfigGetShardingRolesBody string
 )
 
 // cconfigTopology used to export topology information from a Tarantool

cli/replicaset/lua/cconfig/bootstrap_vshard_body.lua

@@ -1,12 +1,15 @@
+local sharding = require('config'):get().sharding
+if sharding == nil or sharding.roles == nil then
+    error("sharding roles are not configured, please make sure managed cluster is sharded")
+end
+
 local ok, vshard = pcall(require, 'vshard')
 if not ok then
     error("failed to require vshard module")
 end
-local fiber = require('fiber')
-local config = require('config')
 
 local is_router = false
-for _, role in ipairs(config:get().sharding.roles) do
+for _, role in ipairs(sharding.roles) do
     if role == "router" then
         is_router = true
         break
@@ -19,6 +22,7 @@ end
 
 pcall(vshard.router.master_search_wakeup)
 
+local fiber = require('fiber')
 local timeout = ...
 local deadline = fiber.time() + timeout
 local ok, err

cli/replicaset/lua/cconfig/get_sharding_roles_body.lua

@@ -0,0 +1,5 @@
+local sharding = require('config'):get().sharding
+if sharding == nil or sharding.roles == nil then
+    error("sharding roles are not configured, please make sure managed cluster is sharded")
+end
+return sharding.roles

test/conftest.py

@@ -1,3 +1,4 @@
+import itertools
 import os
 import platform
 import shutil
@@ -277,34 +278,41 @@ def tt_app(tt, tt_path, tt_instances, tt_running_targets, tt_post_start):
     yield app
 
 
+# Fixture to be used to avoid port collision in different tarantool instances.
+@pytest.fixture(scope="session")
+def port_factory():
+    # It is expected that step by step existent tests with hardcoded ports
+    # are adapted to use this fixture. It is also planned that the new tests
+    # are able to be run in parallel so for now let's start far beyond the
+    # default 3301 that is widely used in existent tests to avoid collision.
+    ports = itertools.count(5501)
+
+    def _port_factory():
+        return next(ports)
+
+    return _port_factory
+
+
 @pytest.fixture
 def cluster_params():
-    return None
+    return tt_helper.make_cluster_params(dict())
 
 
 @pytest.fixture
-def cluster(request, tt, cluster_params):
+def cluster(request, tt, cluster_params, port_factory):
     if utils.is_tarantool_less_3():
         pytest.skip("centralized config requires Tarantool v3.x")
 
-    params = dict(
-        app_name="cluster_app",
-        num_replicasets=1,
-        num_replicas=3,
-        username="client",
-        password="secret",
-    )
-    if cluster_params is not None:
-        params.update(cluster_params)
-
     input_params = [
-        params["num_replicasets"],
-        params["num_replicas"],
-        params["username"],
-        params["password"],
+        cluster_params["num_replicasets"],
+        cluster_params["num_replicas"],
+        cluster_params["username"],
+        cluster_params["password"],
     ]
-    app = tt_helper.TtCluster(tt, params["app_name"], input_params)
+    app = tt_helper.TtCluster(tt, cluster_params["app_name"], input_params)
     request.addfinalizer(lambda: app.stop("--yes"))
+    app.update_ports(cluster_params["host"], port_factory)
+
     return app
 
 

test/integration/replicaset/test_replicaset_vshard.py

@@ -352,3 +352,19 @@ def test_vshard_bootstrap_not_enough_timeout(tt_cmd, vshard_cconfig_app_timeout_
     assert rc == 1
     assert "failed to bootstrap vshard" in out
     assert "attempt to index field '_configdata_applied' (a nil value)" in out
+
+
+@pytest.mark.parametrize("target_type", ["APP", "INST", "URI"])
+def test_vshard_bootstrap_non_vshard(tt, cluster, target_type):
+    targets = {
+        "APP": cluster.app_name,
+        "INST": f"{cluster.app_name}:{cluster.instances[0]['name']}",
+        "URI": f"client:secret@{cluster.instances[0]['endpoint']}",
+    }
+
+    p = cluster.start()
+    assert p.returncode == 0
+    assert cluster.wait_for_running(5)
+    p = tt.run("rs", "vshard", "bootstrap", targets[target_type])
+    assert p.returncode != 0
+    assert "sharding roles are not configured" in p.stdout

test/tt_helper.py

@@ -118,11 +118,27 @@ def __init__(self, tt, app_name, input_params):
         input = "".join(["\n" if x is None else f"{x}\n" for x in input_params])
         p = self.__tt.run("create", "cluster", "--name", self.__app_name, input=input)
         assert p.returncode == 0
+        self.__instances = None
 
     @property
     def app_name(self):
         return self.__app_name
 
+    @property
+    def instances(self):
+        if self.__instances is None:
+            self.__instances = []
+            for group in self.config["groups"].values():
+                for rs in group["replicasets"].values():
+                    for inst_name, inst in rs["instances"].items():
+                        self.__instances.append(
+                            dict(
+                                name=inst_name,
+                                endpoint=inst["iproto"]["listen"][0]["uri"],
+                            ),
+                        )
+        return self.__instances
+
     @property
     def config_path(self):
         return self.__tt.work_dir / self.__app_name / "config.yaml"
@@ -133,6 +149,7 @@ def config(self):
 
     @config.setter
     def config(self, config):
+        self.__instances = None
         self.config_path.write_text(yaml.dump(config))
 
     def start(self, *args):
@@ -165,6 +182,37 @@ def update_config_leaves(self, other):
         utils.update_dict_leaves(config, other)
         self.config = config
 
+    def update_instances_config(self, configure_instance_func, *args):
+        groups = {}
+        for group_name, group in self.config["groups"].items():
+            replicasets = {}
+            for rs_name, rs in group["replicasets"].items():
+                instances = {}
+                for inst_name in rs["instances"]:
+                    instances[inst_name] = configure_instance_func(*args)
+                replicasets[rs_name] = {"instances": instances}
+            groups[group_name] = {"replicasets": replicasets}
+
+        self.update_config_leaves({"groups": groups})
+
+    def update_ports(self, host, port_factory):
+        def configure_instance_port(host, port_factory):
+            return {"iproto": {"listen": [{"uri": f"{host}:{port_factory()}"}]}}
+
+        self.update_instances_config(configure_instance_port, host, port_factory)
+
+
+def make_cluster_params(params):
+    default_params = dict(
+        app_name="cluster_app",
+        num_replicasets=1,
+        num_replicas=3,
+        username="client",
+        password="secret",
+        host="127.0.0.1",
+    )
+    return default_params | params
+
 
 def status(tt, *args):
     rc, out = tt.exec("status", *args)