hiro/ray
1
0
Fork 0
mirror of https://github.com/vale981/ray synced 2025-03-08 11:31:40 -05:00
Code Issues Projects Releases Packages Wiki Activity Actions

[metrics] Clean up object manager stats (#10316)

Browse source
This commit is contained in:
Edward Oakes 2020-08-26 13:43:06 -05:00 • committed by GitHub
parent dc378a80b7
commit c35ad8237d
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 40 additions and 21 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
src/ray/object_manager/object_manager.cc
View file

@ -118,6 +118,7 @@ void ObjectManager::HandleObjectAdded(
RAY_LOG(DEBUG) << "Object added " << object_id; RAY_LOG(DEBUG) << "Object added " << object_id;
RAY_CHECK(local_objects_.count(object_id) == 0); RAY_CHECK(local_objects_.count(object_id) == 0);
local_objects_[object_id].object_info = object_info; local_objects_[object_id].object_info = object_info;
used_memory_ += object_info.data_size + object_info.metadata_size;
ray::Status status = ray::Status status =
object_directory_->ReportObjectAdded(object_id, self_node_id_, object_info); object_directory_->ReportObjectAdded(object_id, self_node_id_, object_info);
@ -146,6 +147,8 @@ void ObjectManager::NotifyDirectoryObjectDeleted(const ObjectID &object_id) {
RAY_CHECK(it != local_objects_.end()); RAY_CHECK(it != local_objects_.end());
auto object_info = it->second.object_info; auto object_info = it->second.object_info;
local_objects_.erase(it); local_objects_.erase(it);
used_memory_ -= object_info.data_size + object_info.metadata_size;
RAY_CHECK(!local_objects_.empty() || used_memory_ == 0);
ray::Status status = ray::Status status =
object_directory_->ReportObjectRemoved(object_id, self_node_id_, object_info); object_directory_->ReportObjectRemoved(object_id, self_node_id_, object_info);
} }
@ -902,24 +905,13 @@ std::string ObjectManager::DebugString() const {
} }
void ObjectManager::RecordMetrics() const { void ObjectManager::RecordMetrics() const {
int64_t used_memory = 0; stats::ObjectStoreAvailableMemory().Record(config_.object_store_memory - used_memory_);
for (const auto &it : local_objects_) { stats::ObjectStoreUsedMemory().Record(used_memory_);
object_manager::protocol::ObjectInfoT object_info = it.second.object_info; stats::ObjectStoreLocalObjects().Record(local_objects_.size());
used_memory += object_info.data_size + object_info.metadata_size; stats::ObjectManagerWaitRequests().Record(active_wait_requests_.size());
} stats::ObjectManagerPullRequests().Record(pull_requests_.size());
stats::ObjectManagerStats().Record(used_memory, stats::ObjectManagerUnfulfilledPushRequests().Record(unfulfilled_push_requests_.size());
{{stats::ValueTypeKey, "used_object_store_memory"}}); stats::ObjectManagerProfileEvents().Record(profile_events_.size());
stats::ObjectManagerStats().Record(local_objects_.size(),
{{stats::ValueTypeKey, "num_local_objects"}});
stats::ObjectManagerStats().Record(active_wait_requests_.size(),
{{stats::ValueTypeKey, "num_active_wait_requests"}});
stats::ObjectManagerStats().Record(
unfulfilled_push_requests_.size(),
{{stats::ValueTypeKey, "num_unfulfilled_push_requests"}});
stats::ObjectManagerStats().Record(pull_requests_.size(),
{{stats::ValueTypeKey, "num_pull_requests"}});
stats::ObjectManagerStats().Record(profile_events_.size(),
{{stats::ValueTypeKey, "num_profile_events"}});
} }
} // namespace ray } // namespace ray

3
src/ray/object_manager/object_manager.h
View file

@ -470,6 +470,9 @@ class ObjectManager : public ObjectManagerInterface,
/// Client id - object manager gRPC client. /// Client id - object manager gRPC client.
std::unordered_map<ClientID, std::shared_ptr<rpc::ObjectManagerClient>> std::unordered_map<ClientID, std::shared_ptr<rpc::ObjectManagerClient>>
remote_object_manager_clients_; remote_object_manager_clients_;
/// Running sum of the amount of memory used in the object store.
int64_t used_memory_ = 0;
}; };
} // namespace ray } // namespace ray

30
src/ray/stats/metric_defs.h
View file

@ -61,9 +61,33 @@ static Gauge RestartingActors("restarting_actors", "Number of restarting actors.
static Gauge DeadActors("dead_actors", "Number of dead actors.", "actors"); static Gauge DeadActors("dead_actors", "Number of dead actors.", "actors");
static Gauge ObjectManagerStats("object_manager_stats", static Gauge ObjectStoreAvailableMemory(
"Stat the metric values of object in raylet", "pcs", "object_store_available_memory",
{ValueTypeKey}); "Amount of memory currently available in the object store.", "bytes");
static Gauge ObjectStoreUsedMemory(
"object_store_used_memory",
"Amount of memory currently occupied in the object store.", "bytes");
static Gauge ObjectStoreLocalObjects("object_store_num_local_objects",
"Number of objects currently in the object store.",
"objects");
static Gauge ObjectManagerWaitRequests("object_manager_num_wait_requests",
"Number of pending wait requests for objects.",
"requests");
static Gauge ObjectManagerPullRequests("object_manager_num_pull_requests",
"Number of active pull requests for objects.",
"requests");
static Gauge ObjectManagerUnfulfilledPushRequests(
"object_manager_unfulfilled_push_requests",
"Number of unfulfilled push requests for objects.", "requests");
static Gauge ObjectManagerProfileEvents("object_manager_num_buffered_profile_events",
"Number of locally-buffered profile events.",
"events");
static Gauge TaskDependencyManagerStats("task_dependency_manager_stats", static Gauge TaskDependencyManagerStats("task_dependency_manager_stats",
"Stat the metric values of task dependency.", "Stat the metric values of task dependency.",