Retries on 502

thebeeland · web-flow · commit 1b0a13613e36 · 2021-07-12T14:42:51.000-04:00
Implements retries with ascending backoff on 502 errors.
diff --git a/shotgun_api3/shotgun.py b/shotgun_api3/shotgun.py
@@ -3322,17 +3322,39 @@ def _call_rpc(self, method, params, include_auth_params=True, first=False):
         if self.config.localized is True:
             req_headers["locale"] = "auto"
 
-        http_status, resp_headers, body = self._make_call("POST", self.config.api_path,
-                                                          encoded_payload, req_headers)
-        LOG.debug("Completed rpc call to %s" % (method))
-        try:
-            self._parse_http_status(http_status)
-        except ProtocolError as e:
-            e.headers = resp_headers
-            # 403 is returned with custom error page when api access is blocked
-            if e.errcode == 403:
-                e.errmsg += ": %s" % body
-            raise
+        attempt = 1
+        max_attempts = 4 # Three retries on failure
+        backoff = 0.75 # Seconds to wait before retry, times the attempt number
+
+        while attempt <= max_attempts:
+            http_status, resp_headers, body = self._make_call(
+                "POST",
+                self.config.api_path,
+                encoded_payload,
+                req_headers,
+            )
+
+            LOG.debug("Completed rpc call to %s" % (method))
+
+            try:
+                self._parse_http_status(http_status)
+            except ProtocolError as e:
+                e.headers = resp_headers
+
+                # We've seen some rare instances of SG returning 502 for issues that
+                # appear to be caused by something internal to SG. We're going to
+                # allow for limited retries for those specifically.
+                if attempt != max_attempts and e.errcode == 502:
+                    LOG.debug("Got a 502 response. Waiting and retrying...")
+                    time.sleep(float(attempt) * backoff)
+                    attempt += 1
+                    continue
+                elif e.errcode == 403:
+                    # 403 is returned with custom error page when api access is blocked
+                    e.errmsg += ": %s" % body
+                raise
+            else:
+                break
 
         response = self._decode_response(resp_headers, body)
         self._response_errors(response)
diff --git a/tests/test_client.py b/tests/test_client.py
@@ -371,6 +371,13 @@ def test_call_rpc(self):
         expected = "rpc response with list result"
         self.assertEqual(d["results"], rv, expected)
 
+        # Test that we raise on a 502. This is ensuring the retries behavior
+        # in place specific to 502 responses still eventually ends up raising.
+        d = {"results": ["foo", "bar"]}
+        a = {"some": "args"}
+        self._mock_http(d, status=(502, "bad gateway"))
+        self.assertRaises(api.ProtocolError, self.sg._call_rpc, "list", a)
+
     def test_transform_data(self):
         """Outbound data is transformed"""
         timestamp = time.time()
