int Do(CWorkerNodeJobContext& context)
{
CNcbiIstream& is = context.GetIStream();
string input_type;
is >> input_type;
if (input_type != "doubles") {
context.CommitJobWithFailure(
"This worker node can only process the 'doubles' input type.");
return 1;
}
int vsize;
is >> vsize;
vector<double> v(vsize);
for (int i = 0; i < vsize; ++i)
is >> v[i];
unsigned delay = m_SleepTimeDistr.GetNextValue();
if (delay > 0)
SleepMilliSec(delay);
if (m_Random.GetRand() <
TParam_FailureRate::GetDefault() * m_Random.GetMax())
context.CommitJobWithFailure("FAILED");
else {
sort(v.begin(), v.end());
CNcbiOstream& os = context.GetOStream();
os << vsize << ' ';
for (int i = 0; i < vsize; ++i)
os << v[i] << ' ';
context.CommitJob();
}
return 0;
}
void CWNJobWatcher::Notify(const CWorkerNodeJobContext& job_context,
EEvent event)
{
switch (event) {
case eJobStarted:
{
CMutexGuard guard(m_ActiveJobsMutex);
m_ActiveJobs[const_cast<CWorkerNodeJobContext*>(&job_context)] =
SJobActivity();
++m_JobsStarted;
if (m_MaxJobsAllowed > 0 && m_JobsStarted > m_MaxJobsAllowed - 1) {
LOG_POST_X(1, "The maximum number of allowed jobs (" <<
m_MaxJobsAllowed << ") has been reached. "
"Sending the shutdown request." );
CGridGlobals::GetInstance().
RequestShutdown(CNetScheduleAdmin::eNormalShutdown);
}
}
break;
case eJobStopped:
{
CMutexGuard guard(m_ActiveJobsMutex);
m_ActiveJobs.erase(
const_cast<CWorkerNodeJobContext*>(&job_context));
}
break;
case eJobFailed:
++m_JobsFailed;
if (m_MaxFailuresAllowed > 0 &&
m_JobsFailed > m_MaxFailuresAllowed - 1) {
LOG_POST_X(2, "The maximum number of failed jobs (" <<
m_MaxFailuresAllowed << ") has been reached. "
"Shutting down..." );
CGridGlobals::GetInstance().
RequestShutdown(CNetScheduleAdmin::eShutdownImmediate);
}
break;
case eJobSucceeded:
++m_JobsSucceeded;
break;
case eJobReturned:
++m_JobsReturned;
break;
case eJobRescheduled:
++m_JobsRescheduled;
break;
case eJobCanceled:
++m_JobsCanceled;
break;
case eJobLost:
++m_JobsLost;
break;
}
if (event != eJobStarted) {
CGridWorkerNode worker_node(job_context.GetWorkerNode());
Uint8 total_memory_limit = worker_node.GetTotalMemoryLimit();
if (total_memory_limit > 0) { // memory check requested
size_t memory_usage;
if (!GetMemoryUsage(&memory_usage, 0, 0)) {
ERR_POST("Could not check self memory usage" );
} else if (memory_usage > total_memory_limit) {
ERR_POST(Warning << "Memory usage (" << memory_usage <<
") is above the configured limit (" <<
total_memory_limit << ")");
CGridGlobals::GetInstance().RequestShutdown(
CNetScheduleAdmin::eNormalShutdown,
RESOURCE_OVERUSE_EXIT_CODE);
}
}
int total_time_limit = worker_node.GetTotalTimeLimit();
if (total_time_limit > 0 && // time check requested
time(0) > worker_node.GetStartupTime() + total_time_limit)
CGridGlobals::GetInstance().RequestShutdown(
CNetScheduleAdmin::eNormalShutdown, RESOURCE_OVERUSE_EXIT_CODE);
}
}
int Do(CWorkerNodeJobContext& context)
{
context.GetCleanupEventSource()->AddListener(
new CSampleJobCleanupListener("Job-do"));
LOG_POST( context.GetJobKey() + " " + context.GetJobInput());
LOG_POST( "This parameter is read from a config file: " << m_Param);
// 1. Get an input data from the client
// (You can use ASN.1 de-serialization here)
//
CNcbiIstream& is = context.GetIStream();
CNcbiOstream& os = context.GetOStream();
string output_type;
is >> output_type; // could be "doubles" or "html"
LOG_POST( "Output type: " << output_type);
int count;
is >> count;
vector<double> dvec;
dvec.reserve(count);
LOG_POST( "Getting " << count << " doubles from stream...");
for (int i = 0; i < count; ++i) {
if (!is.good()) {
ERR_POST( "Input stream error. Index : " << i );
// If anything bad happened, throw an exception
// and its message will be delivered to the client.
throw runtime_error("Worker node input stream error");
}
// Don't forget to check if shutdown has been requested
if (count % 1000 == 0)
switch (context.GetShutdownLevel()) {
case CNetScheduleAdmin::eShutdownImmediate:
case CNetScheduleAdmin::eDie:
// Either this job is not needed anymore (canceled,
// expired or already executed elsewhere), or the
// server is going down and the job's execution
// should be gracefully (yet urgently) aborted and
// the job returned back to the network queue
// for execution by other worker node instances.
context.ReturnJob();
return 1;
default:
break;
}
double d;
is >> d;
dvec.push_back(d);
}
// 2. Doing some time consuming job here
// Well behaved algorithm checks from time to time if
// immediate shutdown has been requested and gracefully return
// without calling context.CommitJob()
//
for (int i = 0; i < m_Iters; ++i) {
switch (context.GetShutdownLevel()) {
case CNetScheduleAdmin::eShutdownImmediate:
case CNetScheduleAdmin::eDie:
// Either this job is not needed anymore (canceled,
// expired or already executed elsewhere), or the
// server is going down and the job's execution
// should be gracefully (yet urgently) aborted and
// the job returned back to the network queue
// for execution by other worker node instances.
context.ReturnJob();
return 1;
default:
break;
}
context.PutProgressMessage("Iteration " + NStr::IntToString(i+1) +
" from " + NStr::IntToString(m_Iters));
SleepSec(m_SleepSec);
}
sort(dvec.begin(), dvec.end());
// 3. Return the result to the client
// (You can use ASN.1 serialization here)
//
// CNcbiOstream& os = context.GetOStream();
if (output_type == "html")
os << "<html><head><title>"
"Sample Grid Worker Result Page"
"</title></head><body>"
"<p>Sample Grid Worker Result</p>";
else
os << dvec.size() << ' ';
for (int i = 0; i < count; ++i) {
if (!os.good()) {
ERR_POST( "Output stream error. Index : " << i );
throw runtime_error("Worker node output stream error");
}
os << dvec[i] << ' ';
}
if (output_type == "html")
os << "</body></html>";
// 4. Indicate that the job is done and the result
// can be delivered to the client.
//
context.CommitJob();
LOG_POST( "Job " << context.GetJobKey() << " is done.");
return 0;
}
