parse setup command errors for more precise tracebacks

Author: jlewitt1

runhouse/resources/hardware/cluster.py

@@ -39,6 +39,7 @@
     generate_default_name,
     install_conda,
     locate_working_dir,
+    parse_traceback_error,
     run_command_with_password_login,
     run_setup_command,
     thread_coroutine,
@@ -1231,9 +1232,14 @@ def _start_or_restart_helper(
             stream_logs=True,
             node=self.head_ip,
         )
-
-        if not status_codes[0] == 0:
-            raise ValueError(f"Failed to restart server {self.name}")
+        ret_code = status_codes[0]
+        if ret_code != 0:
+            raise RuntimeError(
+                parse_traceback_error(
+                    status_codes[1],
+                    fallback_msg=f"Failed to restart server {self.name}",
+                )
+            )
 
         if https_flag:
             rns_address = self.rns_address or self.name

runhouse/resources/packages/package.py

@@ -22,6 +22,7 @@
     install_conda,
     is_python_package_string,
     locate_working_dir,
+    parse_traceback_error,
     run_setup_command,
 )
 
@@ -281,22 +282,31 @@ def _install(
                 conda_env_name=conda_env_name, cluster=cluster
             )
             logger.info(f"Running via install_method pip: {install_cmd}")
-            retcode = run_setup_command(install_cmd, cluster=cluster, node=node)[0]
-            if retcode != 0:
+            status_codes = run_setup_command(install_cmd, cluster=cluster, node=node)
+            ret_code = status_codes[0]
+            if ret_code != 0:
                 raise RuntimeError(
-                    f"Pip install {install_cmd} failed, check that the package exists and is available for your platform."
+                    parse_traceback_error(
+                        status_codes[1],
+                        fallback_msg=f"Pip install {install_cmd} failed, check that the package exists "
+                        f"and is available for your platform.",
+                    )
                 )
 
         elif self.install_method == "conda":
             install_cmd = self._conda_install_cmd(
                 conda_env_name=conda_env_name, cluster=cluster
             )
             logger.info(f"Running via install_method conda: {install_cmd}")
-            retcode = run_setup_command(install_cmd, cluster=cluster, node=node)[0]
-            if retcode != 0:
+            status_codes = run_setup_command(install_cmd, cluster=cluster, node=node)
+            ret_code = status_codes[0]
+            if ret_code != 0:
                 raise RuntimeError(
-                    f"Conda install {install_cmd} failed, check that the package exists and is "
-                    "available for your platform."
+                    parse_traceback_error(
+                        status_codes[1],
+                        fallback_msg=f"Conda install {install_cmd} failed, check that the package exists and is "
+                        "available for your platform.",
+                    )
                 )
 
         elif self.install_method == "reqs":
@@ -305,11 +315,19 @@ def _install(
             )
             if install_cmd:
                 logger.info(f"Running via install_method reqs: {install_cmd}")
-                retcode = run_setup_command(install_cmd, cluster=cluster, node=node)[0]
-                if retcode != 0:
+                status_codes = run_setup_command(
+                    install_cmd, cluster=cluster, node=node
+                )
+                ret_code = status_codes[0]
+                if ret_code != 0:
                     raise RuntimeError(
-                        f"Reqs install {install_cmd} failed, check that the package exists and is available for your platform."
+                        parse_traceback_error(
+                            status_codes[1],
+                            fallback_msg=f"Reqs install {install_cmd} failed, check that the package exists "
+                            f"and is available for your platform.",
+                        )
                     )
+
             else:
                 logger.info(
                     f"{self.install_target.full_local_path_str()}/requirements.txt not found, skipping reqs install"

runhouse/utils.py

@@ -86,6 +86,18 @@ def install_conda(cluster: "Cluster" = None, node: Optional[str] = None):
             raise RuntimeError("Could not install Conda.")
 
 
+def parse_traceback_error(traceback_msg: str, fallback_msg: str):
+    traceback_match = re.search(
+        r"Traceback.*?\n(.+Error: .+)", traceback_msg, re.DOTALL
+    )
+    if not traceback_match:
+        return fallback_msg
+
+    error_message = traceback_match.group(1).strip()
+    cleaned_message = re.sub(r"\x1b\[[0-9;]*m", "", error_message)
+    return cleaned_message
+
+
 def create_conda_env_on_cluster(
     conda_env_name: str,
     conda_config: Dict,