christoph/fail
1
Fork 0
Code Actions Packages Releases Activity

Added reconnect-mechanism in JobClient with random backoff time.

Browse Source 
git-svn-id: https://www4.informatik.uni-erlangen.de/i4svn/danceos/trunk/devel/fail@1001 8c4709b5-6ec9-48aa-a5cd-a96041d1645a
This commit is contained in:
adrian
2012-03-16 14:10:18 +00:00
parent face74ef9a
commit 5a12ab5f3c
2 changed files with 46 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

66
core/jobserver/JobClient.cc
View File

@ -1,39 +1,55 @@
#include "JobClient.hpp" #include "JobClient.hpp"
namespace fi { namespace fi {
JobClient::JobClient(std::string server, int port) JobClient::JobClient(std::string server, int port)
{ {
m_server_port = port; m_server_port = port;
m_server = server; m_server = server;
m_server_ent = gethostbyname(m_server.c_str()); m_server_ent = gethostbyname(m_server.c_str());
if(m_server_ent == NULL){ if(m_server_ent == NULL) {
perror("Cannot resolve host."); perror("[Client@gethostbyname()]");
exit(1); exit(1);
} }
srand(time(NULL));
} }
bool JobClient::connectToServer(){ bool JobClient::connectToServer()
// Connect to server {
struct sockaddr_in serv_addr; // Connect to server
m_sockfd = socket(AF_INET, SOCK_STREAM, 0); struct sockaddr_in serv_addr;
if(m_sockfd < 0){ m_sockfd = socket(AF_INET, SOCK_STREAM, 0);
perror("socket"); if(m_sockfd < 0) {
exit(0); perror("[Client@socket()]");
} exit(0);
}
/* Enable address reuse */ /* Enable address reuse */
int on = 1; int on = 1;
setsockopt( m_sockfd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on) ); setsockopt( m_sockfd, SOL_SOCKET, SO_REUSEADDR, &on, sizeof(on) );
memset(&serv_addr, 0, sizeof(serv_addr)); memset(&serv_addr, 0, sizeof(serv_addr));
serv_addr.sin_family = AF_INET; serv_addr.sin_family = AF_INET;
memcpy(&serv_addr.sin_addr.s_addr, m_server_ent->h_addr, m_server_ent->h_length); memcpy(&serv_addr.sin_addr.s_addr, m_server_ent->h_addr, m_server_ent->h_length);
serv_addr.sin_port = htons(m_server_port); serv_addr.sin_port = htons(m_server_port);
if (connect(m_sockfd, (sockaddr *)&serv_addr, sizeof(serv_addr)) < 0) {
perror("connect()");
return false;
}
int retries = 3;
while(true) {
if(connect(m_sockfd, (sockaddr *)&serv_addr, sizeof(serv_addr)) < 0) {
perror("[Client@connect()]");
if(retries > 0) {
int delay = rand() % 5 + 3;
cout << "[Client] Retrying to connect to server in " << delay << "s..." << endl;
sleep(delay);
--retries;
continue;
}
return false; // finally: unable to connect, give it up :-(
}
break; // connected! :-)
}
cout << "[Client] Connected established!" << endl;
return true; return true;
} }
@ -89,7 +105,7 @@ bool JobClient::sendResult(ExperimentData& result)
ctrlmsg.set_command(FailControlMessage_Command_RESULT_FOLLOWS); ctrlmsg.set_command(FailControlMessage_Command_RESULT_FOLLOWS);
ctrlmsg.set_build_id(42); ctrlmsg.set_build_id(42);
ctrlmsg.set_workloadid(result.getWorkloadID()); ctrlmsg.set_workloadid(result.getWorkloadID());
cout << "Sending back result [" << std::dec << result.getWorkloadID() << "]" << endl; cout << "[Client] Sending back result [" << std::dec << result.getWorkloadID() << "]..." << endl;
SocketComm::send_msg(m_sockfd, ctrlmsg); SocketComm::send_msg(m_sockfd, ctrlmsg);
SocketComm::send_msg(m_sockfd, result.getMessage()); SocketComm::send_msg(m_sockfd, result.getMessage());
// close connection. // close connection.

6
core/jobserver/JobClient.hpp
View File

@ -8,8 +8,12 @@
#ifndef __JOB_CLIENT_H__ #ifndef __JOB_CLIENT_H__
#define __JOB_CLIENT_H__ #define __JOB_CLIENT_H__
#include "SocketComm.hpp"
#include <string> #include <string>
#include <ctime>
#include <stdlib.h>
#include <unistd.h>
#include <iostream>
#include "SocketComm.hpp"
#include "controller/ExperimentData.hpp" #include "controller/ExperimentData.hpp"
#include "jobserver/messagedefs/FailControlMessage.pb.h" #include "jobserver/messagedefs/FailControlMessage.pb.h"