Added caller for REPORT_ADDR to backend.py

lib/backend.py

@@ -66,6 +66,9 @@ 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()
@@ -104,23 +107,19 @@ class Backend:
 
     #######################################Private#######################################
     def _fetch_pubkey(self):
-        command = ["curl", "-X", "GET", "https://run.vast.ai/pubkey/"]
+        report_addr = self.report_addr.rstrip("/")
-        result = subprocess.check_output(command, universal_newlines=True)
+        command = ["curl", "-X", "GET", f"{report_addr}/pubkey/"]
-        log.debug("public key:")
+        try:
-        log.debug(result)
+            result = subprocess.check_output(command, universal_newlines=True)
-        key = None
+            log.debug("public key:")
-        for _ in range(5):
+            log.debug(result)
-            try:
+            key = RSA.import_key(result)
-                key = RSA.import_key(result)
+            if key is not None:
-                break
+                return key
-            except ValueError as e:
+        except (ValueError , subprocess.CalledProcessError) as e:
-                log.debug(f"Error downloading key: {e}")
+            log.debug(f"Error downloading key: {e}")
-                time.sleep(15)
+        self.backend_errored("Failed to get autoscaler pubkey")
-        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,
@@ -315,10 +314,10 @@ 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()
+                    payload = self.benchmark_handler.make_benchmark_payload()
-                    # _ = await self.__call_api(
+                    _ = await self.__call_api(
-                    #     handler=self.benchmark_handler, payload=payload
+                        handler=self.benchmark_handler, payload=payload
-                    # )
+                    )
                     return float(f.readline())
             except FileNotFoundError:
                 pass
@@ -393,7 +392,7 @@ 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