Update Signal alert handling

python/signal_alert.py

@@ -5,23 +5,28 @@ from __future__ import annotations
 import json
 import logging
 import socket
-from os import getenv
 
 logger = logging.getLogger(__name__)
 
 
-def signal_alert(body: str) -> None:
-    """Send a signal alert.
+class SignalRPCError(RuntimeError):
+    """signal-cli returned an error or closed without responding."""
+
+
+def signal_alert(body: str, from_phone: str, to_phone: str, *, timeout: float = 4.0) -> None:
+    """Send a Signal alert through the local signal-cli daemon.
 
     Args:
-        body (str): The body of the alert.
-    """
-    from_phone = getenv("SIGNAL_ALERT_FROM_PHONE")
-    to_phone = getenv("SIGNAL_ALERT_TO_PHONE")
-    if not from_phone or not to_phone:
-        logger.info("SIGNAL_ALERT_FROM_PHONE or SIGNAL_ALERT_TO_PHONE not set")
-        return
+        body: The body of the alert.
+        from_phone: The Signal account sending the alert.
+        to_phone: The Signal account receiving the alert.
+        timeout: Seconds to wait on each socket operation.
 
+    Raises:
+        SignalRPCError: If signal-cli returns an error or closes the
+            connection before responding.
+        OSError: If the socket is unreachable or an operation times out.
+    """
     signal_rpc_id = "signal-alert"
     request = {
         "jsonrpc": "2.0",
@@ -36,7 +41,7 @@ def signal_alert(body: str) -> None:
 
     try:
         with socket.socket(socket.AF_UNIX, socket.SOCK_STREAM) as client:
-            client.settimeout(4.0)
+            client.settimeout(timeout)
             client.connect("/run/signal-cli/socket")
             client.sendall(json.dumps(request).encode() + b"\n")
 
@@ -46,11 +51,14 @@ def signal_alert(body: str) -> None:
                 while b"\n" in buffer:
                     line, buffer = buffer.split(b"\n", maxsplit=1)
                     response = json.loads(line)
-                    if response.get("id") == signal_rpc_id:
-                        if error := response.get("error"):
-                            logger.error("Signal RPC returned an error: %s", error)
-                        return
-
-            logger.error("Signal RPC socket closed before returning a response")
-    except OSError, json.JSONDecodeError:
-        logger.exception("Unable to communicate with signal-cli")
+                    if response.get("id") != signal_rpc_id:
+                        continue
+                    if rpc_error := response.get("error"):
+                        error = f"Signal RPC returned an error: {rpc_error}"
+                        raise SignalRPCError(error)
+                    return
+            error = "Signal RPC socket closed before returning a response"
+            raise SignalRPCError(error)
+    except (OSError, json.JSONDecodeError) as exc:
+        error = f"Signal RPC failed: {exc!r}"
+        raise SignalRPCError(error) from exc

python/tools/snapshot_manager.py

@@ -6,6 +6,7 @@ import logging
 import sys
 import tomllib
 from functools import cache
+from os import getenv
 from pathlib import Path  # noqa: TC003 This is required for the typer CLI
 from re import compile as re_compile
 from re import search
@@ -24,6 +25,15 @@ def main(config_file: Path) -> None:
     configure_logger(level="DEBUG")
     logger.info("Starting snapshot_manager")
 
+    from_phone = getenv("SIGNAL_ALERT_FROM_PHONE")
+    to_phone = getenv("SIGNAL_ALERT_TO_PHONE")
+    if not from_phone:
+        error = "SIGNAL_ALERT_FROM_PHONE environment variable is not set"
+        raise ValueError(error)
+    if not to_phone:
+        error = "SIGNAL_ALERT_TO_PHONE environment variable is not set"
+        raise ValueError(error)
+
     try:
         time_stamp = get_time_stamp()
 
@@ -33,14 +43,14 @@ def main(config_file: Path) -> None:
             if status != "snapshot created":
                 msg = f"{dataset.name} failed to create snapshot {time_stamp}"
                 logger.error(msg)
-                signal_alert(msg)
+                signal_alert(msg, from_phone, to_phone)
                 continue
             count_lookup = get_count_lookup(config_file, dataset.name)
             logger.info(f"using {count_lookup} for {dataset.name}")
-            get_snapshots_to_delete(dataset, count_lookup)
+            get_snapshots_to_delete(dataset, count_lookup, from_phone, to_phone)
     except Exception:
         logger.exception("snapshot_manager failed")
-        signal_alert("snapshot_manager failed")
+        signal_alert("snapshot_manager failed", from_phone, to_phone)
         sys.exit(1)
     else:
         logger.info("snapshot_manager completed")
@@ -92,12 +102,16 @@ def load_config_data(config_file: Path) -> dict[str, dict[str, int]]:
 def get_snapshots_to_delete(
     dataset: Dataset,
     count_lookup: dict[str, int],
+    from_phone: str,
+    to_phone: str,
 ) -> None:
     """Get snapshots to delete.
 
     Args:
         dataset (Dataset): the dataset
         count_lookup (dict[str, int]): the count lookup
+        from_phone (str): The Signal account sending alerts.
+        to_phone (str): The Signal account receiving alerts.
     """
     snapshots = dataset.get_snapshots()
 
@@ -127,7 +141,7 @@ def get_snapshots_to_delete(
         for snapshot in snapshots_being_deleted:
             if error := dataset.delete_snapshot(snapshot):
                 error_message = f"{dataset.name}@{snapshot} failed to delete: {error}"
-                signal_alert(error_message)
+                signal_alert(error_message, from_phone, to_phone)
                 logger.error(error_message)