Change memcache_socket_timeout to a float

It's not unusual to set 0.25 as a socket_timeout value, and this was
not possible. This can have a dramatic impact for performance when one
memcached server is down, especially when keystone is involved.

Change-Id: I01356b7873ad78cb38779f5c5ddd2b932889c911
Partial-Bug: #1731921

oslo_cache/_opts.py

@@ -76,11 +76,11 @@ FILE_OPTIONS = {
                    help='Number of seconds memcached server is considered dead'
                    ' before it is tried again. (dogpile.cache.memcache and'
                    ' oslo_cache.memcache_pool backends only).'),
-        cfg.IntOpt('memcache_socket_timeout',
-                   default=3,
-                   help='Timeout in seconds for every call to a server.'
-                   ' (dogpile.cache.memcache and oslo_cache.memcache_pool'
-                   ' backends only).'),
+        cfg.FloatOpt('memcache_socket_timeout',
+                     default=3.0,
+                     help='Timeout in seconds for every call to a server.'
+                     ' (dogpile.cache.memcache and oslo_cache.memcache_pool'
+                     ' backends only).'),
         cfg.IntOpt('memcache_pool_maxsize',
                    default=10,
                    help='Max total number of open connections to every'

oslo_cache/backends/memcache_pool.py

@@ -45,7 +45,7 @@ class PooledMemcachedBackend(memcached_backend.MemcachedBackend):
             self.url,
             arguments={
                 'dead_retry': arguments.get('dead_retry', 5 * 60),
-                'socket_timeout': arguments.get('socket_timeout', 3),
+                'socket_timeout': arguments.get('socket_timeout', 3.0),
             },
             maxsize=arguments.get('pool_maxsize', 10),
             unused_timeout=arguments.get('pool_unused_timeout', 60),

releasenotes/notes/memcache_socket_timeout-a7db772f052c107e.yaml

@@ -0,0 +1,5 @@
+---
+features:
+  - |
+    `memcache_socket_timeout` is changed to float, given the same default value
+    '3.0'.