Don't set retries for restore snapshot

With this commit we use the standard retry handling from the
Elasticsearch client. Previously, we've implemented a special logic to
add a `retries` parameter but it is ineffective and instead is passed a
request parameter to Elasticsearch.

Relates elastic#799

esrally/driver/runner.py

@@ -1158,15 +1158,6 @@ class RestoreSnapshot(Runner):
     """
     def __call__(self, es, params):
         request_params = params.get("request-params", {})
-        if "retries" not in request_params:
-            # It is possible that there is a proxy in between the cluster and our client which has a shorter timeout
-            # configured. In that case, the proxy would return 504, the client would retry the operation and hit an
-            # error about an index that is already existing. This error message is confusing and thus we explicitly
-            # disallow the client to ever retry.
-            #
-            # see also the docs for ``retries`` in the ``urllib3.connectionpool.urlopen``.
-            request_params["retries"] = 0
-
         es.snapshot.restore(repository=mandatory(params, "repository", repr(self)),
                             snapshot=mandatory(params, "snapshot", repr(self)),
                             body=params.get("body"),

tests/driver/runner_test.py

@@ -2070,10 +2070,7 @@ def test_restore_snapshot_with_body(self, es):
                                                         }
                                                     },
                                                     wait_for_completion=True,
-                                                    params={
-                                                        "request_timeout": 7200,
-                                                        "retries": 0
-                                                    })
+                                                    params={"request_timeout": 7200})
 
 
 class IndicesRecoveryTests(TestCase):