diff options
| author |  Derek Murray <mrry@google.com> | 2018-04-18 15:27:05 -0700 |
|---|---|---|
| committer |  TensorFlower Gardener <gardener@tensorflow.org> | 2018-04-18 15:30:46 -0700 |
| commit | 80f60ea37ed77b3dbe1d983f101a5efba2fd4f2e (patch) | |
| tree | c9a5679baed7ff915eb78ac853f02c62224e4b39 /tensorflow/core/distributed_runtime/worker_session.cc | |
| parent | e662c3fcfcd03fd091b032a5a33971428f4cdb89 (diff) | |
Never use the LegacySession when a Master explicitly calls CreateWorkerSession.
Previously, if the session handle was unrecognized by the worker, it
would default to using the LegacySession. This prevents us from
noticing that a server has been restarted.
To address the problem in a backwards-compatible way, we add a bit to
each session-handle-carrying worker request, indicating whether the
master believes that CreateWorkerSession has been called. If this bit
is set and the handle is unrecognized, the worker will raise an
AbortedError, which can be caught by high-level frameworks such as
`tf.estimator`.
Note that CreateWorkerSession is not yet used by default, and a
follow-up change will add that.
PiperOrigin-RevId: 193427057
Diffstat (limited to 'tensorflow/core/distributed_runtime/worker_session.cc')
| -rw-r--r-- | tensorflow/core/distributed_runtime/worker_session.cc | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/tensorflow/core/distributed_runtime/worker_session.cc b/tensorflow/core/distributed_runtime/worker_session.cc index cb7059b36e..18886babd5 100644 --- a/tensorflow/core/distributed_runtime/worker_session.cc +++ b/tensorflow/core/distributed_runtime/worker_session.cc @@ -97,6 +97,7 @@ WorkerSession::WorkerSession(const string& session_name, worker_cache(new WorkerFreeListCache(std::move(worker_cache))), device_mgr(std::move(device_mgr)), graph_mgr(std::move(graph_mgr)), - cluster_flr(new ClusterFunctionLibraryRuntime(this)) {} + cluster_flr( + new ClusterFunctionLibraryRuntime(this, !session_name.empty())) {} } // namespace tensorflow |