backend: respect timeout on socket/timeout errors pu

While retrying idempotent requests (even on timeouts) would
prevent stale sockets from being noticed, it is better to kill
the socket and immediately propagate the timeout error to the
user.  Retrying in a timeout may cause a request/response to
take longer (perhaps _much_ longer) than the timeout configured
by the user.

1 files changed, 4 insertions, 3 deletions

diff --git a/lib/mogilefs/backend.rb b/lib/mogilefs/backend.rb
index cbeec3d..afc2f1a 100644
--- a/lib/mogilefs/backend.rb
+++ b/lib/mogilefs/backend.rb
@@ -246,15 +246,16 @@ class MogileFS::Backend
           raise EOFError, "end of file reached after: #{request.inspect}"
         # fall through to retry in loop
       rescue SystemCallError,
-             MogileFS::UnreadableSocketError,
-             MogileFS::InvalidResponseError, # truncated response
-             MogileFS::Timeout
+             MogileFS::InvalidResponseError # truncated response
         # we got a successful timed_write, but not a timed_gets
         if idempotent
           failed = true
+          shutdown_unlocked(false)
           retry
         end
         shutdown_unlocked(true)
+      rescue MogileFS::UnreadableSocketError, MogileFS::Timeout
+        shutdown_unlocked(true)
       rescue
         # we DO NOT want the response we timed out waiting for, to crop up later
         # on, on the same socket, intersperesed with a subsequent request!  we