void CondorResource::DoScheddPoll()
{
int rc;
ScheddPollInfo *poll_info = NULL;
if ( ( registeredJobs.IsEmpty() || resourceDown ) &&
scheddStatusActive == false ) {
// No jobs or we can't talk to the schedd, so no point
// in polling
daemonCore->Reset_Timer( scheddPollTid, BatchStatusInterval() );
return;
}
if ( gahp->isStarted() == false ) {
// The gahp isn't started yet. Wait a few seconds for a CondorJob
// object to start it (and possibly initialize x509 credentials).
daemonCore->Reset_Timer( scheddPollTid, 5 );
return;
}
PollInfoByName.lookup( HashKey( HashName( scheddName, poolName, NULL ) ),
poll_info );
daemonCore->Reset_Timer( scheddPollTid, TIMER_NEVER );
if ( scheddStatusActive == false ) {
// We share polls across all CondorResource objects going to
// the same schedd. If another object has done a poll
// recently, then don't bother doing one ourselves.
if ( poll_info == NULL ) {
poll_info = new ScheddPollInfo;
poll_info->m_lastPoll = 0;
poll_info->m_pollActive = false;
PollInfoByName.insert( HashKey( HashName( scheddName, poolName,
NULL ) ),
poll_info );
}
if ( poll_info->m_pollActive == true ||
poll_info->m_lastPoll + BatchStatusInterval() > time(NULL) ) {
daemonCore->Reset_Timer( scheddPollTid, BatchStatusInterval() );
return;
}
// start schedd status command
dprintf( D_FULLDEBUG, "Starting collective poll: %s\n",
scheddName );
std::string constraint;
// create a list of jobs we expect to hear about in our
// status command
// Since we're sharing the results of this status command with
// all CondorResource objects going to the same schedd, look
// for their jobs as well.
poll_info->m_submittedJobs.Rewind();
while ( poll_info->m_submittedJobs.Next() ) {
poll_info->m_submittedJobs.DeleteCurrent();
}
CondorResource *next_resource;
BaseJob *job;
std::string job_id;
ResourcesByName.startIterations();
while ( ResourcesByName.iterate( next_resource ) != 0 ) {
if ( strcmp( scheddName, next_resource->scheddName ) ||
strcmp( poolName ? poolName : "",
next_resource->poolName ? next_resource->poolName : "" ) ) {
continue;
}
next_resource->registeredJobs.Rewind();
while ( ( job = next_resource->registeredJobs.Next() ) ) {
if ( job->jobAd->LookupString( ATTR_GRID_JOB_ID, job_id ) ) {
poll_info->m_submittedJobs.Append( (CondorJob *)job );
}
}
}
formatstr( constraint, "(%s)", submitter_constraint.c_str() );
rc = gahp->condor_job_status_constrained( scheddName,
constraint.c_str(),
NULL, NULL );
if ( rc != GAHPCLIENT_COMMAND_PENDING ) {
dprintf( D_ALWAYS,
"gahp->condor_job_status_constrained returned %d for remote schedd: %s\n",
rc, scheddName );
EXCEPT( "condor_job_status_constrained failed!" );
}
scheddStatusActive = true;
poll_info->m_pollActive = true;
} else {
// finish schedd status command
int num_status_ads;
ClassAd **status_ads = NULL;
ASSERT( poll_info );
rc = gahp->condor_job_status_constrained( NULL, NULL,
&num_status_ads,
&status_ads );
if ( rc == GAHPCLIENT_COMMAND_PENDING ) {
return;
} else if ( rc != 0 ) {
dprintf( D_ALWAYS,
"gahp->condor_job_status_constrained returned %d for remote schedd %s\n",
rc, scheddName );
dprintf( D_ALWAYS, "Requesting ping of resource\n" );
RequestPing( NULL );
}
if ( rc == 0 ) {
for ( int i = 0; i < num_status_ads; i++ ) {
int cluster, proc;
int rc2;
std::string job_id_string;
BaseJob *base_job = NULL;
CondorJob *job;
if( status_ads[i] == NULL ) {
dprintf(D_ALWAYS, "DoScheddPoll was given null pointer for classad #%d\n", i);
continue;
}
status_ads[i]->LookupInteger( ATTR_CLUSTER_ID, cluster );
status_ads[i]->LookupInteger( ATTR_PROC_ID, proc );
formatstr( job_id_string, "condor %s %s %d.%d", scheddName,
poolName, cluster, proc );
rc2 = BaseJob::JobsByRemoteId.lookup( HashKey( job_id_string.c_str() ),
base_job );
job = dynamic_cast<CondorJob*>( base_job );
if ( rc2 == 0 ) {
job->NotifyNewRemoteStatus( status_ads[i] );
poll_info->m_submittedJobs.Delete( job );
} else {
delete status_ads[i];
}
}
poll_info->m_lastPoll = time(NULL);
}
poll_info->m_pollActive = false;
if ( status_ads != NULL ) {
free( status_ads );
}
// Check if any jobs were missing from the status result
if ( rc == 0 ) {
CondorJob *job;
std::string job_id;
poll_info->m_submittedJobs.Rewind();
while ( ( job = poll_info->m_submittedJobs.Next() ) ) {
if ( job->jobAd->LookupString( ATTR_GRID_JOB_ID, job_id ) ) {
// We should have gotten a status ad for this job,
// but didn't. Tell the job that there may be
// something wrong by giving it a NULL status ad.
job->NotifyNewRemoteStatus( NULL );
}
poll_info->m_submittedJobs.DeleteCurrent();
}
}
scheddStatusActive = false;
dprintf( D_FULLDEBUG, "Collective poll complete: %s\n", scheddName );
daemonCore->Reset_Timer( scheddPollTid, BatchStatusInterval() );
}
}
void NordugridResource::DoJobStatus()
{
if ( ( registeredJobs.IsEmpty() || resourceDown ) &&
m_jobStatusActive == false ) {
// No jobs or we can't talk to the resource, so no point
// in polling
daemonCore->Reset_Timer( m_jobStatusTid, NordugridJob::probeInterval );
return;
}
if ( m_statusGahp->isStarted() == false ) {
// The gahp isn't started yet. Wait a few seconds for a Job
// object to start it and initialize x509 credentials.
daemonCore->Reset_Timer( m_jobStatusTid, 5 );
return;
}
daemonCore->Reset_Timer( m_jobStatusTid, TIMER_NEVER );
StringList results;
if ( m_jobStatusActive == false ) {
// start ldap status command
dprintf( D_FULLDEBUG, "Starting ldap poll: %s\n", resourceName );
std::string ldap_server = resourceName;
size_t pos = ldap_server.find_first_of( ':' );
if ( pos != std::string::npos ) {
ldap_server.erase( pos );
}
std::string filter;
sprintf( filter, "(&(objectclass=nordugrid-job)(nordugrid-job-globalowner=%s))", proxySubject );
int rc = m_statusGahp->nordugrid_ldap_query( ldap_server.c_str(), "mds-vo-name=local,o=grid", filter.c_str(), "nordugrid-job-globalid,nordugrid-job-status",
results );
if ( rc != GAHPCLIENT_COMMAND_PENDING ) {
dprintf( D_ALWAYS,
"gahp->nordugrid_ldap_query returned %d for resource %s\n",
rc, resourceName );
EXCEPT( "nordugrid_ldap_query failed!" );
}
m_jobStatusActive = true;
} else {
// finish ldap status command
int rc = m_statusGahp->nordugrid_ldap_query( NULL, NULL, NULL, NULL, results );
if ( rc == GAHPCLIENT_COMMAND_PENDING ) {
return;
} else if ( rc != 0 ) {
dprintf( D_ALWAYS,
"gahp->nordugrid_ldap_query returned %d for resource %s: %s\n",
rc, resourceName, m_statusGahp->getErrorString() );
dprintf( D_ALWAYS, "Requesting ping of resource\n" );
RequestPing( NULL );
}
if ( rc == 0 ) {
const char *next_job_id = NULL;
const char *next_status = NULL;
const char *next_attr;
std::string key;
results.rewind();
do {
next_attr = results.next();
if ( next_attr != NULL && *next_attr != '\0' ) {
// Save the attributes we're interested in
if ( !strncmp( next_attr, "nordugrid-job-globalid: ", 24 ) ) {
next_job_id = next_attr;
} else if ( !strncmp( next_attr, "nordugrid-job-status: ", 22 ) ) {
next_status = next_attr;
}
continue;
}
// We just reached the end of a record. Process it.
// If we don't have the attributes we expect, skip it.
if ( next_job_id && next_status ) {
int rc2;
NordugridJob *job;
sprintf( key, "nordugrid %s %s", resourceName,
strrchr( next_job_id, '/' ) + 1 );
rc2 = BaseJob::JobsByRemoteId.lookup( HashKey( key.c_str() ),
(BaseJob*&)job );
if ( rc2 == 0 ) {
job->NotifyNewRemoteStatus( strchr( next_status, ' ' ) + 1 );
}
}
next_job_id = NULL;
next_status = NULL;
} while ( next_attr );
}
m_jobStatusActive = false;
dprintf( D_FULLDEBUG, "ldap poll complete: %s\n", resourceName );
daemonCore->Reset_Timer( m_jobStatusTid, NordugridJob::probeInterval );
}
}
