From 1eafdce7dd2e52b81544e5c2bef9e7356b9cc936 Mon Sep 17 00:00:00 2001
From: CamDavidsonPilon <cam.davidson.pilon@gmail.com>
Date: Tue, 3 Sep 2024 20:36:06 -0400
Subject: [PATCH] introduce the watchdog 'fix' too

---
 pioreactor/background_jobs/leader/watchdog.py | 51 +------------------
 1 file changed, 2 insertions(+), 49 deletions(-)

diff --git a/pioreactor/background_jobs/leader/watchdog.py b/pioreactor/background_jobs/leader/watchdog.py
index ac96d2f6..2a57f23a 100644
--- a/pioreactor/background_jobs/leader/watchdog.py
+++ b/pioreactor/background_jobs/leader/watchdog.py
@@ -46,62 +46,15 @@ def announce_new_workers(self) -> None:
                     )
 
     def watch_for_lost_state(self, state_message: MQTTMessage) -> None:
-        # generally, I hate this code below...
-
         unit = state_message.topic.split("/")[1]
 
-        # don't check workers that aren't part of the cluster
-        current_workers = get_workers_in_inventory()
-
         # ignore if leader is "lost"
         if (
             (state_message.payload.decode() == self.LOST)
             and (unit != self.unit)
-            and (unit in current_workers)
+            and (unit in get_workers_in_inventory())
         ):
-            # TODO: this song-and-dance works for monitor, why not extend it to other jobs...
-
-            self.logger.warning(f"{unit} seems to be lost. Trying to re-establish connection...")
-            time.sleep(5)
-
-            if self.state != self.READY:
-                # when the entire Rpi shuts down, ex via sudo reboot, monitor can publish a lost. This code will halt the shutdown.
-                # let's return early.
-                return
-
-            # this is a hack! If the monitor job is in state READY, it will no op any transition.
-            # so we set to sleeping for a second, and the back to ready.
-            self.pub_client.publish(
-                f"pioreactor/{unit}/{UNIVERSAL_EXPERIMENT}/monitor/$state/set", self.SLEEPING
-            )
-            time.sleep(1)
-            self.pub_client.publish(
-                f"pioreactor/{unit}/{UNIVERSAL_EXPERIMENT}/monitor/$state/set", self.READY
-            )
-            ###
-            time.sleep(10)
-
-            if self.state != self.READY:
-                # when the entire Rpi shuts down, ex via sudo reboot, monitor can publish a lost. This code will halt the shutdown.
-                # let's return early.
-                return
-
-            msg = subscribe(  # I don't think this can be self.sub_client because we are in a callback.
-                f"pioreactor/{unit}/{UNIVERSAL_EXPERIMENT}/monitor/$state",
-                timeout=15,
-                name=self.job_name,
-                retries=1,
-            )
-            if msg is None:
-                return
-
-            current_state = msg.payload.decode()
-
-            if current_state == self.LOST:
-                # failed, let's confirm to user
-                self.logger.error(f"{unit} was lost.")
-            else:
-                self.logger.info(f"Update: {unit} is connected. All is well.")
+            self.logger.warning(f"{unit} seems to be lost.")
 
     def start_passive_listeners(self) -> None:
         self.subscribe_and_callback(