fix: handle empty 'Retry-After' header from GitLab

srikanthchelluri · srikanthchelluri · commit 7a3724f3fca9 · 2019-01-08T21:54:09.000-05:00
When requests are throttled (HTTP response code 429), python-gitlab
assumed that 'Retry-After' existed in the response headers. This is
not always the case and so the request fails due to a KeyError. The
change in this commit adds a rudimentary exponential backoff to the
'http_request' method, which defaults to 10 retries but can be set
to -1 to retry without bound.
diff --git a/docs/api-usage.rst b/docs/api-usage.rst
@@ -299,7 +299,9 @@ Rate limits
 
 python-gitlab obeys the rate limit of the GitLab server by default.  On
 receiving a 429 response (Too Many Requests), python-gitlab sleeps for the
-amount of time in the Retry-After header that GitLab sends back.
+amount of time in the Retry-After header that GitLab sends back.  If GitLab
+does not return a response with the Retry-After header, python-gitlab will
+perform an exponential backoff.
 
 If you don't want to wait, you can disable the rate-limiting feature, by
 supplying the ``obey_rate_limit`` argument.
@@ -312,6 +314,18 @@ supplying the ``obey_rate_limit`` argument.
    gl = gitlab.gitlab(url, token, api_version=4)
    gl.projects.list(all=True, obey_rate_limit=False)
 
+If you do not disable the rate-limiting feature, you can supply a custom value
+for ``max_retries``; by default, this is set to 10. To retry without bound when
+throttled, you can set this parameter to -1. This parameter is ignored if
+``obey_rate_limit`` is set to ``False``.
+
+.. code-block:: python
+
+   import gitlab
+   import requests
+
+   gl = gitlab.gitlab(url, token, api_version=4)
+   gl.projects.list(all=True, max_retries=12)
 
 .. warning::
 
diff --git a/gitlab/__init__.py b/gitlab/__init__.py
@@ -477,6 +477,10 @@ def http_request(self, verb, path, query_data={}, post_data=None,
         # obey the rate limit by default
         obey_rate_limit = kwargs.get("obey_rate_limit", True)
 
+        # set max_retries to 10 by default, disable by setting it to -1
+        max_retries = kwargs.get("max_retries", 10)
+        cur_retries = 0
+
         while True:
             result = self.session.send(prepped, timeout=timeout, **settings)
 
@@ -486,9 +490,13 @@ def http_request(self, verb, path, query_data={}, post_data=None,
                 return result
 
             if 429 == result.status_code and obey_rate_limit:
-                wait_time = int(result.headers["Retry-After"])
-                time.sleep(wait_time)
-                continue
+                if max_retries == -1 or cur_retries < max_retries:
+                    wait_time = 2 ** cur_retries * 0.1
+                    if "Retry-After" in result.headers:
+                        wait_time = int(result.headers["Retry-After"])
+                    cur_retries += 1
+                    time.sleep(wait_time)
+                    continue
 
             error_message = result.content
             try:
