Fix flaky test test_many_connections in test_odbc_interaction

alexey-milovidov · claude · alexey-milovidov · commit 366f274ab470 · 2026-03-01T01:13:29.000+01:00
The `odbc-bridge` crashes with SIGSEGV when handling 25 concurrent requests simultaneously (each UNION ALL branch opens a separate HTTP connection to the bridge). The crash occurs in the thread pool when `ParallelFormattingOutputFormat` tries to spawn formatting threads for each concurrent request, causing resource exhaustion. After the crash, remaining connections get "Connection refused" and the query fails. Use `assert_eq_with_retry` (already used elsewhere in this test file) to retry the query. On retry, `startBridgeSync` detects the dead bridge and restarts it. Closes #92521 https://s3.amazonaws.com/clickhouse-test-reports/json.html?PR=93983&sha=fbe65318fd2687eb5640e14df6cd5cba1455cada&name_0=PR&name_1=Integration%20tests%20%28amd_binary%2C%203%2F5%29 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
diff --git a/tests/integration/test_odbc_interaction/test.py b/tests/integration/test_odbc_interaction/test.py
@@ -957,7 +957,7 @@ def test_many_connections(started_cluster):
         query += "SELECT key FROM {t} UNION ALL "
     query += "SELECT key FROM {t})"
 
-    assert node1.query(query.format(t="test_pg_table")) == "250\n"
+    assert_eq_with_retry(node1, query.format(t="test_pg_table"), "250")
     cursor.execute("DROP TABLE clickhouse.test_pg_table")
 
 
