Mirror/FairMQ
1
0
Fork 0
mirror of https://github.com/FairRootGroup/FairMQ.git synced 2025-10-13 16:46:47 +00:00
Code Issues Packages Projects Releases Wiki Activity

SDK: Improve error handling in case state-change fails on a device

Browse Source 
Replace the log message with

1. Nothing, if the device is already in the target state
2. Abort and call the completion callback with error otherwise
This commit is contained in:
Dennis Klein 2019-07-28 12:00:46 +02:00 committed by Dennis Klein
parent fd282fa950
commit 7b773cde51
2 changed files with 24 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
fairmq/sdk/Topology.cxx
View File

@ -98,7 +98,16 @@ Topology::Topology(DDSTopology topo, DDSSession session)
LOG(error) << "state-changes-unsubscription failed with return code: " << parts[2];
}
} else if (parts[1] == "could not queue") {
LOG(warn) << "Could not queue " << parts[2] << " transition on " << senderId;
std::unique_lock<std::mutex> lock(fMtx);
if (fStateChangeOngoing) {
if (fState.at(fDDSSession.GetTaskId(senderId)).state != fTargetState) {
fStateChangeError =
tools::ToString("Could not queue ", parts[2], " transition on ", senderId);
lock.unlock();
fCV.notify_one();
}
}
}
});
fDDSSession.StartDDSService();
@ -124,13 +133,13 @@ auto Topology::ChangeState(TopologyTransition transition, ChangeStateCallback cb
std::unique_lock<std::mutex> lock(fMtx);
if (fStateChangeOngoing) {
throw std::runtime_error("A state change request is already in progress, concurrent requests are currently not supported");
lock.unlock();
}
LOG(debug) << "Initiating ChangeState with " << transition << " to " << expectedState.at(transition);
fStateChangeOngoing = true;
fChangeStateCallback = cb;
fStateChangeTimeout = timeout;
fTargetState = expectedState.at(transition);
fStateChangeError.clear();
fDDSSession.SendCommand(GetTransitionName(transition));
}
@ -157,12 +166,15 @@ void Topology::WaitForState()
while (!fShutdown) {
if (fStateChangeOngoing) {
try {
std::unique_lock<std::mutex> lock(fMtx);
auto condition = [&] {
// LOG(info) << "checking condition";
// LOG(info) << "fShutdown: " << fShutdown;
// LOG(info) << "condition: " << std::all_of(fState.cbegin(), fState.cend(),
// [&](TopologyState::value_type i) { return i.second.state == fTargetState; });
return fShutdown
|| !fStateChangeError.empty()
|| std::all_of(
fState.cbegin(), fState.cend(), [&](TopologyState::value_type i) {
// TODO Check, if we can make sure that EXITING state change event are not missed
@ -172,8 +184,6 @@ void Topology::WaitForState()
});
};
std::unique_lock<std::mutex> lock(fMtx);
if (fStateChangeTimeout > std::chrono::milliseconds(0)) {
if (!fCV.wait_for(lock, fStateChangeTimeout, condition)) {
// LOG(debug) << "timeout";
@ -189,6 +199,15 @@ void Topology::WaitForState()
}
fStateChangeOngoing = false;
if (!fStateChangeError.empty()) {
TopologyState state = fState;
lock.unlock();
fChangeStateCallback(
{{AsyncOpResultCode::Error, fStateChangeError}, std::move(state)});
break;
}
if (fShutdown) {
LOG(debug) << "Aborting because a shutdown was requested";
TopologyState state = fState;

1
fairmq/sdk/Topology.h
View File

@ -150,6 +150,7 @@ class Topology
ChangeStateCallback fChangeStateCallback;
std::chrono::milliseconds fStateChangeTimeout;
bool fShutdown;
std::string fStateChangeError;
void WaitForState();
void AddNewStateEntry(uint64_t senderId, const std::string& state);