remove redis pubsub from pyworker

lib/backend.py

@@ -66,9 +66,6 @@ class Backend:
     unsecured: bool = dataclasses.field(
         default_factory=lambda: bool(strtobool(os.environ.get("UNSECURED", "false"))),
     )
-    report_addr: str = dataclasses.field(
-        default_factory=lambda: os.environ.get("REPORT_ADDR", "https://run.vast.ai")
-    )
 
     def __post_init__(self):
         self.metrics = Metrics()
@@ -107,19 +104,23 @@ class Backend:
 
     #######################################Private#######################################
     def _fetch_pubkey(self):
-        report_addr = self.report_addr.rstrip("/")
-        command = ["curl", "-X", "GET", f"{report_addr}/pubkey/"]
-        try:
-            result = subprocess.check_output(command, universal_newlines=True)
-            log.debug("public key:")
-            log.debug(result)
-            key = RSA.import_key(result)
-            if key is not None:
-                return key
-        except (ValueError , subprocess.CalledProcessError) as e:
-            log.debug(f"Error downloading key: {e}")
-        self.backend_errored("Failed to get autoscaler pubkey")
-       
+        command = ["curl", "-X", "GET", "https://run.vast.ai/pubkey/"]
+        result = subprocess.check_output(command, universal_newlines=True)
+        log.debug("public key:")
+        log.debug(result)
+        key = None
+        for _ in range(5):
+            try:
+                key = RSA.import_key(result)
+                break
+            except ValueError as e:
+                log.debug(f"Error downloading key: {e}")
+                time.sleep(15)
+        if key is None:
+            self._total_pubkey_fetch_errors += 1
+            if self._total_pubkey_fetch_errors >= MAX_PUBKEY_FETCH_ATTEMPTS:
+                self.backend_errored("Failed to get autoscaler pubkey")
+        return key
 
     async def __handle_request(
         self,
@@ -314,10 +315,10 @@ class Backend:
                 with open(BENCHMARK_INDICATOR_FILE, "r") as f:
                     log.debug("already ran benchmark")
                     # trigger model load
-                    # payload = self.benchmark_handler.make_benchmark_payload()
-                    # _ = await self.__call_api(
-                    #     handler=self.benchmark_handler, payload=payload
-                    # )
+                    payload = self.benchmark_handler.make_benchmark_payload()
+                    _ = await self.__call_api(
+                        handler=self.benchmark_handler, payload=payload
+                    )
                     return float(f.readline())
             except FileNotFoundError:
                 pass
@@ -392,7 +393,7 @@ class Backend:
                         )
                         # some backends need a few seconds after logging successful startup before
                         # they can begin accepting requests
-                        # await sleep(5)
+                        await sleep(5)
                         try:
                             max_throughput = await run_benchmark()
                             self.__start_healthcheck = True

workers/comfyui-json/client.py

@@ -98,7 +98,6 @@ def call_text2image_workflow(
         endpoint=route_response["endpoint"],
         reqnum=route_response["reqnum"],
         url=route_response["url"],
-        request_idx=route_response["request_idx"],
     )
     
     # Build the payload for the worker request

workers/comfyui/client.py

@@ -82,7 +82,6 @@ def call_custom_workflow_for_sd3(
         endpoint=message["endpoint"],
         reqnum=message["reqnum"],
         url=message["url"],
-        request_idx=message["request_idx"],
     )
     workflow = {
         "3": {

workers/openai/server.py

@@ -43,7 +43,6 @@ backend = Backend(
             for error_msg in MODEL_SERVER_ERROR_LOG_MSGS
         ],
     ],
-    max_wait_time=600
 )
 
 

workers/tgi/server.py

@@ -113,7 +113,6 @@ backend = Backend(
             for error_msg in MODEL_SERVER_ERROR_LOG_MSGS
         ],
     ],
-    max_wait_time=600
 )