Skip to content

Automate browser based workflows with AI

shu/revert transaction pooler (#4339)

Shuchang Zheng71db86cbf0be
diff --git a/skyvern/config.py b/skyvern/config.py
index 9c6655566..b9ec07608 100644
--- a/skyvern/config.py
+++ b/skyvern/config.py
@@ -54,7 +54,6 @@ class Settings(BaseSettings):
LONG_RUNNING_TASK_WARNING_RATIO: float = 0.95
MAX_RETRIES_PER_STEP: int = 5
DEBUG_MODE: bool = False
- # Database settings
DATABASE_STRING: str = (
"postgresql+asyncpg://skyvern@localhost/skyvern"
if platform.system() == "Windows"
@@ -63,8 +62,6 @@ class Settings(BaseSettings):
DATABASE_REPLICA_STRING: str | None = None
DATABASE_STATEMENT_TIMEOUT_MS: int = 60000
DISABLE_CONNECTION_POOL: bool = False
- DB_DISABLE_PREPARED_STATEMENTS: bool = False
-
PROMPT_ACTION_HISTORY_WINDOW: int = 1
TASK_RESPONSE_ACTION_SCREENSHOT_COUNT: int = 3
diff --git a/skyvern/forge/sdk/db/agent_db.py b/skyvern/forge/sdk/db/agent_db.py
index 0f1df5358..1405a1a86 100644
--- a/skyvern/forge/sdk/db/agent_db.py
+++ b/skyvern/forge/sdk/db/agent_db.py
@@ -3,23 +3,7 @@ from datetime import datetime, timedelta
from typing import Any, List, Literal, Sequence, overload
import structlog
-from sqlalchemy import (
- Connection,
- and_,
- asc,
- case,
- delete,
- distinct,
- event,
- exists,
- func,
- or_,
- pool,
- select,
- tuple_,
- update,
-)
-from sqlalchemy.engine.url import make_url
+from sqlalchemy import and_, asc, case, delete, distinct, exists, func, or_, pool, select, tuple_, update
from sqlalchemy.exc import (
SQLAlchemyError,
)
@@ -172,58 +156,25 @@ def _serialize_proxy_location(proxy_location: ProxyLocationInput) -> str | None:
return result
-def _connect_args_for_driver(database_string: str) -> dict[str, Any]:
- driver = make_url(database_string).drivername # "postgresql+psycopg" or "postgresql+asyncpg"
- args: dict[str, Any] = {}
-
- if settings.DB_DISABLE_PREPARED_STATEMENTS:
- if driver == "postgresql+psycopg":
- # psycopg3: completely disable prepared statements (None vs 0)
- # 0 disables caching but still uses prepared statements
- # None completely disables prepared statement usage
- args["prepare_threshold"] = None
- elif driver == "postgresql+asyncpg":
- # asyncpg: disable statement cache (prepared statements)
- args["statement_cache_size"] = 0
- else:
- LOG.warning(
- "The database driver might not be well optimized or supported by skyvern: {driver}", driver=driver
- )
-
- return args
-
-
-def _install_statement_timeout(engine: AsyncEngine, timeout_ms: int) -> None:
- if not timeout_ms or timeout_ms <= 0:
- return
+DB_CONNECT_ARGS: dict[str, Any] = {}
- timeout_value = int(timeout_ms)
-
- # Works for direct AND poolers because it's not a startup parameter.
- # Applies per-transaction, which is the most reliable behavior with transaction pooling.
- @event.listens_for(engine.sync_engine, "begin")
- def _set_timeout(conn: Connection) -> None:
- # Use a unique comment with object id to prevent psycopg3 from reusing prepared statement names,
- # which can cause "prepared statement already exists" errors with connection poolers.
- sql = f"SET LOCAL statement_timeout = {timeout_value} /* {id(conn)} */"
- conn.exec_driver_sql(sql)
-
-
-def make_async_engine(database_string: str) -> AsyncEngine:
- engine = create_async_engine(
- database_string,
- json_serializer=_custom_json_serializer,
- connect_args=_connect_args_for_driver(database_string),
- poolclass=pool.NullPool if settings.DISABLE_CONNECTION_POOL else None,
- )
-
- _install_statement_timeout(engine, settings.DATABASE_STATEMENT_TIMEOUT_MS)
- return engine
+if "postgresql+psycopg" in settings.DATABASE_STRING:
+ DB_CONNECT_ARGS = {"options": f"-c statement_timeout={settings.DATABASE_STATEMENT_TIMEOUT_MS}"}
+elif "postgresql+asyncpg" in settings.DATABASE_STRING:
+ DB_CONNECT_ARGS = {"server_settings": {"statement_timeout": str(settings.DATABASE_STATEMENT_TIMEOUT_MS)}}
class AgentDB(BaseAlchemyDB):
def __init__(self, database_string: str, debug_enabled: bool = False, db_engine: AsyncEngine | None = None) -> None:
- super().__init__(db_engine or make_async_engine(database_string))
+ super().__init__(
+ db_engine
+ or create_async_engine(
+ database_string,
+ json_serializer=_custom_json_serializer,
+ connect_args=DB_CONNECT_ARGS,
+ poolclass=pool.NullPool if settings.DISABLE_CONNECTION_POOL else None,
+ )
+ )
self.debug_enabled = debug_enabled
def is_retryable_error(self, error: SQLAlchemyError) -> bool: