Merge pull request #72637 from ClickHouse/startup-scripts-execution-state

add `StartupScriptsExecutionState` metric

programs/server/Server.cpp

@@ -296,6 +296,7 @@ namespace CurrentMetrics
     extern const Metric MergesMutationsMemoryTracking;
     extern const Metric MaxDDLEntryID;
     extern const Metric MaxPushedDDLEntryID;
+    extern const Metric StartupScriptsExecutionState;
 }
 
 namespace ProfileEvents
@@ -366,6 +367,14 @@ namespace ErrorCodes
 }
 
 
+enum StartupScriptsExecutionState : CurrentMetrics::Value
+{
+    NotFinished = 0,
+    Success = 1,
+    Failure = 2,
+};
+
+
 static std::string getCanonicalPath(std::string && path)
 {
     Poco::trimInPlace(path);
@@ -782,9 +791,12 @@ void loadStartupScripts(const Poco::Util::AbstractConfiguration & config, Contex
             startup_context->makeQueryContext();
             executeQuery(read_buffer, write_buffer, true, startup_context, callback, QueryFlags{ .internal = true }, std::nullopt, {});
         }
+
+        CurrentMetrics::set(CurrentMetrics::StartupScriptsExecutionState, StartupScriptsExecutionState::Success);
     }
     catch (...)
     {
+        CurrentMetrics::set(CurrentMetrics::StartupScriptsExecutionState, StartupScriptsExecutionState::Failure);
         tryLogCurrentException(log, "Failed to parse startup scripts file");
     }
 }

src/Common/CurrentMetrics.cpp

@@ -356,6 +356,8 @@
     M(SharedDatabaseCatalogTablesInLocalDropDetachQueue, "Number of tables in the queue for local drop or detach in Shared Catalog.") \
     \
     M(MergeTreeIndexGranularityInternalArraysTotalSize, "The total size of all internal arrays in Merge Tree index granularity objects in bytes.") \
+    \
+    M(StartupScriptsExecutionState, "State of startup scripts execution: 0 = not finished, 1 = success, 2 = failure.") \
 
 #ifdef APPLY_FOR_EXTERNAL_METRICS
     #define APPLY_FOR_METRICS(M) APPLY_FOR_BUILTIN_METRICS(M) APPLY_FOR_EXTERNAL_METRICS(M)

tests/integration/test_startup_scripts_execution_state/config/bad_script.xml

@@ -0,0 +1,10 @@
+<clickhouse>
+    <startup_scripts>
+        <scripts>
+            <query>SELECT 42;</query>
+        </scripts>
+        <scripts>
+            <query>SELECT * FROM non_existent_table;</query>
+        </scripts>
+    </startup_scripts>
+</clickhouse>

tests/integration/test_startup_scripts_execution_state/config/good_script.xml

@@ -0,0 +1,11 @@
+<clickhouse>
+    <startup_scripts>
+        <scripts>
+            <condition>SELECT 0;</condition>
+            <query>SELECT * FROM non_existent_table;</query>
+        </scripts>
+        <scripts>
+            <query>SELECT 42;</query>
+        </scripts>
+    </startup_scripts>
+</clickhouse>

tests/integration/test_startup_scripts_execution_state/config/users.xml

@@ -0,0 +1,8 @@
+<clickhouse>
+    <users>
+        <default>
+            <profile>default</profile>
+            <no_password></no_password>
+        </default>
+    </users>
+</clickhouse>

tests/integration/test_startup_scripts_execution_state/test.py

@@ -0,0 +1,56 @@
+import random
+import string
+import time
+from enum import Enum
+
+import pytest
+
+from helpers.cluster import ClickHouseCluster
+
+cluster = ClickHouseCluster(__file__)
+good = cluster.add_instance(
+    "good",
+    main_configs=["config/users.xml", "config/good_script.xml"],
+    stay_alive=True,
+)
+bad = cluster.add_instance(
+    "bad",
+    main_configs=["config/users.xml", "config/bad_script.xml"],
+    stay_alive=True,
+)
+
+
+@pytest.fixture(scope="module")
+def start_cluster():
+    try:
+        cluster.start()
+        yield cluster
+    finally:
+        cluster.shutdown()
+
+
+def test_startup_execution_state(start_cluster):
+    """
+    Making sure that the StartupScriptsExecutionState metric is set correctly.
+    """
+
+    STATE_SUCCESS = 1
+    STATE_FAILURE = 2
+
+    assert (
+        int(
+            good.query(
+                "SELECT value FROM system.metrics WHERE metric = 'StartupScriptsExecutionState'"
+            ).strip()
+        )
+        == STATE_SUCCESS
+    )
+
+    assert (
+        int(
+            bad.query(
+                "SELECT value FROM system.metrics WHERE metric = 'StartupScriptsExecutionState'"
+            ).strip()
+        )
+        == STATE_FAILURE
+    )