From 882d4f381b8af72d8859b7d849c241d9578e24ff Mon Sep 17 00:00:00 2001
From: Horst Schirmeier <horst.schirmeier@tu-dortmund.de>
Date: Thu, 16 Jan 2014 17:54:50 +0100
Subject: [PATCH] jobclient: bugfix: faster shutdown at campaign end

The JobClient currently waits a LONG time until it really shuts down
after not having reached the server in sendResultsToServer() (which is
unfortunately the by far most probable point in the code to determine
this):

 -  A different bug (fixed in the previous commit) provoked the
    situation that a (way) too large amount of jobs was fetched
    before.
 -  sendResult() (called after each experiment iteration) realized
    that CLIENT_JOB_REQUEST_SEC seconds are over, and tried to
    prematurely call home to send first results (without planning to
    get new jobs yet).
 -  If the server was gone (done, or aborted), connect in
    sendResultsToServer() failed after several retries and timeouts.
 -  All subsequent calls to sendResult() retried connecting to the
    server (again, with retries and timeouts), once for each remaining
    job.
 -  When all jobs were done, getParam() tries to connect a last time,
    finally telling the experiment that nobody's home.

This resulted in client shutdown times of up to four hours (for the
default CLIENT_JOB_LIMIT of 1000) after the campaign server
terminated.  This change solves the issue by not handing out new
(cached) jobs after the connect failed once, making the experiment
terminate quickly.

Change-Id: I0d8cb2e084d783aca74c51a503fa72eb2b2eb0b7
---
 src/core/efw/JobClient.cc  | 24 ++++++++++++++++++++++--
 src/core/efw/JobClient.hpp |  2 ++
 2 files changed, 24 insertions(+), 2 deletions(-)

diff --git a/src/core/efw/JobClient.cc b/src/core/efw/JobClient.cc
index 2bb89f80..34732338 100644
--- a/src/core/efw/JobClient.cc
+++ b/src/core/efw/JobClient.cc
@@ -22,6 +22,7 @@ JobClient::JobClient(const std::string& server, int port)
 	m_job_total = 0;
 	m_job_runtime_total = 0;
 	m_job_throughput = 1; // client gets only one job at the first request
+	m_connect_failed = false;
 }
 
 JobClient::~JobClient()
@@ -32,6 +33,11 @@ JobClient::~JobClient()
 
 bool JobClient::connectToServer()
 {
+	// don't retry server connects to speedup shutdown at campaign end
+	if (m_connect_failed) {
+		return false;
+	}
+
 	int retries = CLIENT_RETRY_COUNT;
 	while (true) {
 		// Connect to server
@@ -69,6 +75,7 @@ bool JobClient::connectToServer()
 			cout << "[Client] Unable to reconnect (tried " << CLIENT_RETRY_COUNT << " times); "
 			     << "I'll give it up!" << endl;
 			     // TODO: Log-level?
+			m_connect_failed = true;
 			return false; // finally: unable to connect, give it up :-(
 		}
 		break; // connected! :-)
@@ -81,6 +88,11 @@ bool JobClient::connectToServer()
 
 bool JobClient::getParam(ExperimentData& exp)
 {
+	// die immediately if a previous connect already failed
+	if (m_connect_failed) {
+		return false;
+	}
+
 	while (1) { // Here we try to acquire a parameter set
 		switch (tryToGetExperimentData(exp)) {
 			// Jobserver will sent workload, params are set in \c exp
@@ -190,10 +202,10 @@ bool JobClient::sendResult(ExperimentData& result)
 			m_job_runtime.reset();
 			m_job_runtime.startTimer();
 			m_job_total += m_results.size();
-			sendResultsToServer();
+			// tell caller whether we failed phoning home
+			return sendResultsToServer();
 		}
 
-		//If there are more jobs for the experiment store result
 		return true;
 	} else {
 		//Stop time measurement and calculate new throughput
@@ -221,6 +233,14 @@ bool JobClient::sendResultsToServer()
 {
 	if (m_results.size() != 0) {
 		if (!connectToServer()) {
+			// clear results, although we didn't get them to safety; otherwise,
+			// subsequent calls to sendResult() may and the destructor will
+			// retry sending them, resulting in a large shutdown time
+			while (m_results.size()) {
+				delete &m_results.front()->getMessage();
+				delete m_results.front();
+				m_results.pop_front();
+			}
 			return false;
 		}
 
diff --git a/src/core/efw/JobClient.hpp b/src/core/efw/JobClient.hpp
index de4153c5..f0d03157 100644
--- a/src/core/efw/JobClient.hpp
+++ b/src/core/efw/JobClient.hpp
@@ -37,6 +37,8 @@ private:
 	std::deque<ExperimentData*> m_parameters;
 	std::deque<ExperimentData*> m_results;
 
+	bool m_connect_failed;
+
 	bool connectToServer();
 	bool sendResultsToServer();
 	FailControlMessage_Command tryToGetExperimentData(ExperimentData& exp);