Adding anvil state to crash report (#1029)

lib/agent0/agent0/base/config/environment_config.py

@@ -25,6 +25,9 @@ class EnvironmentConfig(FrozenClass):
     halt_on_errors: bool = False
     # if halt_on_errors is True, halt_on_slippage controls if we halt when slippage happens
     halt_on_slippage: bool = False
+    # if true, will dump anvil state in crash report
+    # Defaults to false due to size
+    dump_state_on_crash: bool = False
     # optional output filename for logging
     log_filename: str = ".logging/agent0_logs.log"
     # log level; should be in [logging.DEBUG, logging.INFO, logging.WARNING]

lib/agent0/agent0/hyperdrive/crash_report/crash_report.py

@@ -61,11 +61,6 @@ def default(self, o):
         return json.JSONEncoder.default(self, o)
 
 
-def _get_git_revision_hash() -> str:
-    """Helper function for getting commit hash from git."""
-    return subprocess.check_output(["git", "rev-parse", "HEAD"]).decode("ascii").strip()
-
-
 def setup_hyperdrive_crash_report_logging(log_format_string: str | None = None) -> None:
     """Create a new logging file handler with CRITICAL log level for hyperdrive crash reporting.
 
@@ -125,12 +120,16 @@ def log_hyperdrive_crash_report(trade_result: TradeResult, log_level: int | None
                 # NOTE if this crash report happens in a PR that gets squashed,
                 # we loose this hash.
                 ("commit_hash", _get_git_revision_hash()),
+                ("anvil_dump_state", trade_result.anvil_state),
             ]
         ),
         indent=4,
         cls=ExtendedJSONEncoder,
     )
 
+    # TODO the anvil dump state blows up the output, should likely print to stdout
+    # without the state, and log the full crash report to a file. This allows us
+    # to default state dumps to true, while keeping the output sane.
     logging.log(log_level, crash_report_json)
 
 
@@ -189,3 +188,8 @@ def _hyperdrive_trade_obj_to_dict(trade_obj: types.Trade[HyperdriveMarketAction]
 
 def _hyperdrive_agent_to_dict(agent: HyperdriveAgent):
     return {"address": agent.checksum_address, "policy": agent.policy.name}
+
+
+def _get_git_revision_hash() -> str:
+    """Helper function for getting commit hash from git."""
+    return subprocess.check_output(["git", "rev-parse", "HEAD"]).decode("ascii").strip()

lib/agent0/agent0/hyperdrive/exec/run_agents.py

@@ -136,6 +136,7 @@ def run_agents(
             agent_accounts,
             environment_config.halt_on_errors,
             environment_config.halt_on_slippage,
+            environment_config.dump_state_on_crash,
             last_executed_block,
             liquidate,
         )

lib/agent0/agent0/hyperdrive/exec/trade_loop.py

@@ -25,6 +25,7 @@ def trade_if_new_block(
     agent_accounts: list[HyperdriveAgent],
     halt_on_errors: bool,
     halt_on_slippage: bool,
+    dump_state_on_crash: bool,
     last_executed_block: int,
     liquidate: bool,
 ) -> int:
@@ -104,6 +105,8 @@ def trade_if_new_block(
                 if is_slippage:
                     log_hyperdrive_crash_report(trade_result, logging.WARNING)
                 else:
+                    if dump_state_on_crash:
+                        trade_result.anvil_state = _get_anvil_state_dump(hyperdrive.web3)
                     # Defaults to CRITICAL
                     log_hyperdrive_crash_report(trade_result)
 
@@ -304,3 +307,15 @@ def get_wait_for_new_block(web3: Web3) -> bool:
         # do nothing, this will fail for non anvil nodes and we don't care.
         automine = False
     return not automine
+
+
+def _get_anvil_state_dump(web3: Web3) -> str | None:
+    """Helper function for getting anvil dump state"""
+    result: str | None = None
+    try:
+        response = web3.provider.make_request(method=RPCEndpoint("anvil_dumpState"), params=[])
+        result = response.get("result", False)
+    except Exception:  # pylint: disable=broad-exception-caught
+        # do nothing, this is best effort crash reporting
+        pass
+    return result

lib/agent0/agent0/hyperdrive/state/trade_result.py

@@ -42,6 +42,8 @@ class TradeResult:
         The information of the latest checkpoint
     additional_info : dict[str, Any]
         Additional information used for crash reporting
+    anvil_state: str | None
+        The anvil state dump when the exception occurred
     """
 
     status: TradeStatus
@@ -53,3 +55,4 @@ class TradeResult:
     pool_info: dict[str, Any] | None = None
     checkpoint_info: dict[str, Any] | None = None
     additional_info: dict[str, Any] | None = None
+    anvil_state: str | None = None

tests/multi_trade_per_block_test.py

@@ -142,9 +142,10 @@ def test_multi_trade_per_block(
 
         # Build environment config
         env_config = EnvironmentConfig(
-            delete_previous_logs=False,
+            delete_previous_logs=True,
             halt_on_errors=True,
-            log_filename="system_test",
+            dump_state_on_crash=True,
+            log_filename=".logging/multi_trade_per_block_test.log",
             log_level=logging.INFO,
             log_stdout=True,
             random_seed=1234,