bool BaseResource::Invalidate () {
ClassAd ad;
/* Set the correct types */
SetMyTypeName ( ad, GRID_ADTYPE );
/* We only want to invalidate this resource. Using the tuple
(HashName,SchedName,Owner) as unique id. */
std::string line;
formatstr( line,
"((TARGET.%s =?= \"%s\") && (TARGET.%s =?= \"%s\") && "
"(TARGET.%s =?= \"%s\") && (TARGET.%s =?= \"%s\"))",
ATTR_HASH_NAME, GetHashName (),
ATTR_SCHEDD_NAME, ScheddObj->name (),
ATTR_SCHEDD_IP_ADDR, ScheddObj->addr (),
ATTR_OWNER, myUserName );
ad.AssignExpr ( ATTR_REQUIREMENTS, line.c_str() );
ad.Assign( ATTR_HASH_NAME, GetHashName() );
ad.Assign( ATTR_SCHEDD_NAME, ScheddObj->name() );
ad.Assign( ATTR_SCHEDD_IP_ADDR, ScheddObj->addr() );
ad.Assign( ATTR_OWNER, myUserName );
dprintf (
D_FULLDEBUG,
"BaseResource::InvalidateResource: \n%s\n",
line.c_str() );
return daemonCore->sendUpdates( INVALIDATE_GRID_ADS, &ad, NULL, true ) > 0;
}
/* Function : finalizeDelta
* Description: clears and resets all inner structures and data members declared
* in this class only (excluding inherited classes)
*/
void
ReplicatorStateMachine::finalizeDelta( )
{
ClassAd invalidate_ad;
MyString line;
dprintf( D_ALWAYS, "ReplicatorStateMachine::finalizeDelta started\n" );
utilCancelTimer(m_replicationTimerId);
utilCancelTimer(m_versionRequestingTimerId);
utilCancelTimer(m_versionDownloadingTimerId);
utilCancelTimer(m_updateCollectorTimerId);
m_replicationInterval = -1;
m_hadAliveTolerance = -1;
m_maxTransfererLifeTime = -1;
m_newlyJoinedWaitingVersionInterval = -1;
m_lastHadAliveTime = -1;
m_updateInterval = -1;
if ( m_classAd != NULL ) {
delete m_classAd;
m_classAd = NULL;
}
SetMyTypeName( invalidate_ad, QUERY_ADTYPE );
SetTargetTypeName( invalidate_ad, "Replication" );
line.formatstr( "TARGET.%s == \"%s\"", ATTR_NAME, m_name.Value( ) );
invalidate_ad.AssignExpr( ATTR_REQUIREMENTS, line.Value( ) );
invalidate_ad.Assign( ATTR_NAME, m_name.Value() );
daemonCore->sendUpdates( INVALIDATE_ADS_GENERIC, &invalidate_ad, NULL, false );
}
static void
serialize_input( void )
{
char sinful[16 + 10];
int found_eof = 0,
found_error = 0,
empty = 0;
/** Determine if we are using the command-line options, a file,
or standard input as input */
if ( !stdio ) {
/** Contrive a sinful string based on our IP address */
sprintf ( sinful, "<255.255.255.255:1234>" );
/** We were give all the raw data, so we're going to create
a fake machine ad that we will use when invoking the waking
mechanism */
ad = new ClassAd ();
SetMyTypeName ( *ad, STARTD_ADTYPE );
SetTargetTypeName ( *ad, JOB_ADTYPE );
ad->Assign ( ATTR_HARDWARE_ADDRESS, mac );
ad->Assign ( ATTR_SUBNET_MASK, mask );
ad->Assign ( ATTR_MY_ADDRESS, sinful );
ad->Assign ( ATTR_WOL_PORT, port );
} else {
/** Open the machine ad file or read it from stdin */
if ( fn_in && *fn_in ) {
in = safe_fopen_wrapper ( fn_in, "r" );
} else {
in = stdin;
fn_in = "<stdin>";
}
if ( !in ) {
error ( E_FOPEN, fn_in, strerror ( errno ), errno );
}
/** Serialize the machine ad from a file */
ad = new ClassAd (
in,
"?$#%^&*@", /* sufficiently random garbage? */
found_eof,
found_error,
empty );
if ( found_error ) {
error (
E_CLASSAD,
found_error );
}
}
}
void
Defrag::invalidatePublicAd() {
ClassAd invalidate_ad;
std::string line;
SetMyTypeName(invalidate_ad, QUERY_ADTYPE);
SetTargetTypeName(invalidate_ad, "Defrag");
formatstr(line,"%s == \"%s\"", ATTR_NAME, m_daemon_name.c_str());
invalidate_ad.AssignExpr(ATTR_REQUIREMENTS, line.c_str());
invalidate_ad.Assign(ATTR_NAME, m_daemon_name.c_str());
daemonCore->sendUpdates(INVALIDATE_ADS_GENERIC, &invalidate_ad, NULL, false);
}
void
DBMSManager::InvalidatePublicAd() {
ClassAd query_ad;
SetMyTypeName(query_ad, QUERY_ADTYPE);
SetTargetTypeName(query_ad, DBMSD_ADTYPE);
MyString line;
line.formatstr("%s = TARGET.%s == \"%s\"", ATTR_REQUIREMENTS, ATTR_NAME, m_name.Value());
query_ad.Insert(line.Value());
query_ad.Assign(ATTR_NAME,m_name.Value());
m_collectors->sendUpdates(INVALIDATE_ADS_GENERIC, &query_ad, NULL, true);
//TODO/FIXME - delete the ads of the databases we're advertising
return;
}
void
DBMSManager::InitPublicAd() {
m_public_ad = ClassAd();
//TODO: this is the wrong ADTYPE.
//Use the new generic ad type when it becomes available.
//Until then, actual writes to the collector are disabled,
//because it conflicts with the schedd on the same host..
SetMyTypeName( m_public_ad, DBMSD_ADTYPE );
SetTargetTypeName( m_public_ad, "" );
m_public_ad.Assign(ATTR_MACHINE,get_local_fqdn());
m_public_ad.Assign(ATTR_NAME,m_name.Value());
config_fill_ad( &m_public_ad );
}
int
LogNewClassAd::Play(void *data_structure)
{
int result;
ClassAdHashTable *table = (ClassAdHashTable *)data_structure;
ClassAd *ad = new ClassAd();
SetMyTypeName(*ad, mytype);
SetTargetTypeName(*ad, targettype);
ad->EnableDirtyTracking();
result = table->insert(HashKey(key), ad);
#if defined(HAVE_DLOPEN)
ClassAdLogPluginManager::NewClassAd(key);
#endif
return result;
}
void
Defrag::publish(ClassAd *ad)
{
char *valid_name = build_valid_daemon_name(m_defrag_name.c_str());
ASSERT( valid_name );
m_daemon_name = valid_name;
delete [] valid_name;
SetMyTypeName(*ad, "Defrag");
SetTargetTypeName(*ad, "");
ad->Assign(ATTR_NAME,m_daemon_name.c_str());
m_stats.Tick();
m_stats.Publish(*ad);
daemonCore->publish(ad);
}
void
ReplicatorStateMachine::initializeClassAd()
{
if( m_classAd != NULL) {
delete m_classAd;
m_classAd = NULL;
}
m_classAd = new ClassAd();
SetMyTypeName(*m_classAd, "Replication");
SetTargetTypeName(*m_classAd, "");
m_name.formatstr( "replication@%s -p %d", get_local_fqdn().Value(),
daemonCore->InfoCommandPort( ) );
m_classAd->Assign( ATTR_NAME, m_name.Value( ) );
m_classAd->Assign( ATTR_MY_ADDRESS,
daemonCore->InfoCommandSinfulString( ) );
// publish list of replication nodes
char* buffer = param( "REPLICATION_LIST" );
if ( NULL == buffer ) {
EXCEPT( "ReplicatorStateMachine: No replication list!!" );
}
char* replAddress = NULL;
StringList replList;
MyString attrReplList;
MyString comma;
replList.initializeFromString( buffer );
replList.rewind( );
while( ( replAddress = replList.next() ) ) {
attrReplList += comma;
attrReplList += replAddress;
comma = ",";
}
m_classAd->Assign( ATTR_REPLICATION_LIST, attrReplList.Value( ) );
// publish DC attributes
daemonCore->publish(m_classAd);
free(buffer);
}
void
invalidate_ads() {
ClassAd cmd_ad;
SetMyTypeName( cmd_ad, QUERY_ADTYPE );
SetTargetTypeName( cmd_ad, MASTER_ADTYPE );
MyString line;
std::string escaped_name;
char* default_name = ::strnewp(MasterName);
if(!default_name) {
default_name = default_daemon_name();
}
EscapeAdStringValue( default_name, escaped_name );
line.formatstr( "( TARGET.%s == \"%s\" )", ATTR_NAME, escaped_name.c_str() );
cmd_ad.AssignExpr( ATTR_REQUIREMENTS, line.Value() );
cmd_ad.Assign( ATTR_NAME, default_name );
cmd_ad.Assign( ATTR_MY_ADDRESS, daemonCore->publicNetworkIpAddr());
daemonCore->sendUpdates( INVALIDATE_MASTER_ADS, &cmd_ad, NULL, false );
delete [] default_name;
}
bool BaseResource::SendUpdate () {
ClassAd ad;
/* Set the correct types */
SetMyTypeName ( ad, GRID_ADTYPE );
/* populate class ad with the relevant resource information */
PublishResourceAd ( &ad );
daemonCore->publish( &ad );
std::string tmp;
sPrintAd ( tmp, ad );
dprintf (
D_FULLDEBUG,
"BaseResource::UpdateResource: \n%s\n",
tmp.c_str() );
return daemonCore->sendUpdates( UPDATE_GRID_AD, &ad, NULL, true ) > 0;
}
ClassAd * ProcAd::dumpToAd( piPTR pi ) {
char line[128];
ClassAd *ad = new ClassAd;
SetMyTypeName( *ad, "PROCESS_INFORMATION" );
SetTargetTypeName( *ad, "ENQUIRING_MINDS_WANT_TO_KNOW" );
sprintf ( line, "THIS_PID = %d", pi->pid );
ad->Insert(line);
sprintf ( line, "PARENT_PID = %ld", (long)pi->ppid );
ad->Insert(line);
sprintf ( line, "IMAGE_SIZE = %ld", (long)pi->imgsize );
ad->Insert(line);
sprintf ( line, "RESIDENT_SET_SIZE = %ld", (long)pi->rssize );
ad->Insert(line);
#if HAVE_PSS
if( pi->pssize_available ) {
sprintf ( line, "PROPORTIONAL_SET_SIZE = %ld", (long)pi->pssize );
ad->Insert(line);
}
#endif
sprintf ( line, "MAJOR_PAGE_FAULTS = %ld", (long)pi->majfault );
ad->Insert(line);
sprintf ( line, "MINOR_PAGE_FAULTS = %ld", (long)pi->minfault );
ad->Insert(line);
sprintf ( line, "USER_TIME = %ld", (long)pi->user_time );
ad->Insert(line);
sprintf ( line, "SYSTEM_TIME = %ld", (long)pi->sys_time );
ad->Insert(line);
sprintf ( line, "PROCESS_AGE = %ld", (long)pi->age );
ad->Insert(line);
sprintf ( line, "PERCENT_CPU_USAGE = %6.2f", pi->cpuusage );
ad->Insert(line);
return ad;
}
void
init_classad()
{
if( ad ) delete( ad );
ad = new ClassAd();
SetMyTypeName(*ad, MASTER_ADTYPE);
SetTargetTypeName(*ad, "");
if (MasterName) {
ad->Assign(ATTR_NAME, MasterName);
} else {
char* default_name = default_daemon_name();
if( ! default_name ) {
EXCEPT( "default_daemon_name() returned NULL" );
}
ad->Assign(ATTR_NAME, default_name);
delete [] default_name;
}
#if !defined(WIN32)
ad->Assign(ATTR_REAL_UID, (int)getuid());
#endif
}
QueryResult CondorQuery::
getQueryAd (ClassAd &queryAd)
{
QueryResult result;
ExprTree *tree;
queryAd = extraAttrs;
if (resultLimit > 0) { queryAd.Assign(ATTR_LIMIT_RESULTS, resultLimit); }
result = (QueryResult) query.makeQuery (tree);
if (result != Q_OK) return result;
queryAd.Insert(ATTR_REQUIREMENTS, tree);
// fix types
SetMyTypeName (queryAd, QUERY_ADTYPE);
switch (queryType) {
case DEFRAG_AD:
SetTargetTypeName(queryAd, DEFRAG_ADTYPE);
break;
case STARTD_AD:
case STARTD_PVT_AD:
SetTargetTypeName (queryAd, STARTD_ADTYPE);
break;
case SCHEDD_AD:
SetTargetTypeName (queryAd, SCHEDD_ADTYPE);
break;
case SUBMITTOR_AD:
SetTargetTypeName (queryAd, SUBMITTER_ADTYPE);
break;
case LICENSE_AD:
SetTargetTypeName (queryAd, LICENSE_ADTYPE);
break;
case MASTER_AD:
SetTargetTypeName (queryAd, MASTER_ADTYPE);
break;
case CKPT_SRVR_AD:
SetTargetTypeName (queryAd, CKPT_SRVR_ADTYPE);
break;
case COLLECTOR_AD:
SetTargetTypeName (queryAd, COLLECTOR_ADTYPE);
break;
case NEGOTIATOR_AD:
SetTargetTypeName (queryAd, NEGOTIATOR_ADTYPE);
break;
case STORAGE_AD:
SetTargetTypeName (queryAd, STORAGE_ADTYPE);
break;
case CREDD_AD:
SetTargetTypeName (queryAd, CREDD_ADTYPE);
break;
case GENERIC_AD:
if ( genericQueryType ) {
SetTargetTypeName (queryAd, genericQueryType);
} else {
SetTargetTypeName (queryAd, GENERIC_ADTYPE);
}
break;
case ANY_AD:
SetTargetTypeName (queryAd, ANY_ADTYPE);
break;
case DATABASE_AD:
SetTargetTypeName (queryAd, DATABASE_ADTYPE);
break;
case TT_AD:
SetTargetTypeName (queryAd, TT_ADTYPE);
break;
case GRID_AD:
SetTargetTypeName (queryAd, GRID_ADTYPE);
break;
case HAD_AD:
SetTargetTypeName (queryAd, HAD_ADTYPE);
break;
case ACCOUNTING_AD:
SetTargetTypeName(queryAd, ACCOUNTING_ADTYPE);
break;
default:
return Q_INVALID_QUERY;
}
return Q_OK;
}
QueryResult CondorQuery::
getQueryAd (ClassAd &queryAd)
{
QueryResult result;
ExprTree *tree;
queryAd = extraAttrs;
result = (QueryResult) query.makeQuery (tree);
if (result != Q_OK) return result;
queryAd.Insert(ATTR_REQUIREMENTS, tree);
// fix types
SetMyTypeName (queryAd, QUERY_ADTYPE);
switch (queryType) {
#ifdef HAVE_EXT_POSTGRESQL
case QUILL_AD:
SetTargetTypeName (queryAd, QUILL_ADTYPE);
break;
#endif /* HAVE_EXT_POSTGRESQL */
case DEFRAG_AD:
SetTargetTypeName(queryAd, DEFRAG_ADTYPE);
break;
case STARTD_AD:
case STARTD_PVT_AD:
SetTargetTypeName (queryAd, STARTD_ADTYPE);
break;
case SCHEDD_AD:
SetTargetTypeName (queryAd, SCHEDD_ADTYPE);
break;
case SUBMITTOR_AD:
SetTargetTypeName (queryAd, SUBMITTER_ADTYPE);
break;
case LICENSE_AD:
SetTargetTypeName (queryAd, LICENSE_ADTYPE);
break;
case MASTER_AD:
SetTargetTypeName (queryAd, MASTER_ADTYPE);
break;
case CKPT_SRVR_AD:
SetTargetTypeName (queryAd, CKPT_SRVR_ADTYPE);
break;
case COLLECTOR_AD:
SetTargetTypeName (queryAd, COLLECTOR_ADTYPE);
break;
case NEGOTIATOR_AD:
SetTargetTypeName (queryAd, NEGOTIATOR_ADTYPE);
break;
case STORAGE_AD:
SetTargetTypeName (queryAd, STORAGE_ADTYPE);
break;
case CREDD_AD:
SetTargetTypeName (queryAd, CREDD_ADTYPE);
break;
case GENERIC_AD:
if ( genericQueryType ) {
SetTargetTypeName (queryAd, genericQueryType);
} else {
SetTargetTypeName (queryAd, GENERIC_ADTYPE);
}
break;
case XFER_SERVICE_AD:
SetTargetTypeName (queryAd, XFER_SERVICE_ADTYPE);
break;
case LEASE_MANAGER_AD:
SetTargetTypeName (queryAd, LEASE_MANAGER_ADTYPE);
break;
case ANY_AD:
SetTargetTypeName (queryAd, ANY_ADTYPE);
break;
case DATABASE_AD:
SetTargetTypeName (queryAd, DATABASE_ADTYPE);
break;
case DBMSD_AD:
SetTargetTypeName (queryAd, DBMSD_ADTYPE);
break;
case TT_AD:
SetTargetTypeName (queryAd, TT_ADTYPE);
break;
case GRID_AD:
SetTargetTypeName (queryAd, GRID_ADTYPE);
break;
case HAD_AD:
SetTargetTypeName (queryAd, HAD_ADTYPE);
break;
default:
return Q_INVALID_QUERY;
}
return Q_OK;
}
/* Utility function to create a generic job ad. The caller can then fill
* in the relevant details.
*/
ClassAd *CreateJobAd( const char *owner, int universe, const char *cmd )
{
ClassAd *job_ad = new ClassAd();
SetMyTypeName(*job_ad, JOB_ADTYPE);
SetTargetTypeName(*job_ad, STARTD_ADTYPE);
if ( owner ) {
job_ad->Assign( ATTR_OWNER, owner );
} else {
job_ad->AssignExpr( ATTR_OWNER, "Undefined" );
}
job_ad->Assign( ATTR_JOB_UNIVERSE, universe );
job_ad->Assign( ATTR_JOB_CMD, cmd );
job_ad->Assign( ATTR_Q_DATE, (int)time(NULL) );
job_ad->Assign( ATTR_COMPLETION_DATE, 0 );
job_ad->Assign( ATTR_JOB_REMOTE_WALL_CLOCK, 0.0 );
job_ad->Assign( ATTR_JOB_LOCAL_USER_CPU, 0.0 );
job_ad->Assign( ATTR_JOB_LOCAL_SYS_CPU, 0.0 );
job_ad->Assign( ATTR_JOB_REMOTE_USER_CPU, 0.0 );
job_ad->Assign( ATTR_JOB_REMOTE_SYS_CPU, 0.0 );
// This is a magic cookie, see how condor_submit sets it
job_ad->Assign( ATTR_CORE_SIZE, -1 );
// Are these ones really necessary?
job_ad->Assign( ATTR_JOB_EXIT_STATUS, 0 );
job_ad->Assign( ATTR_ON_EXIT_BY_SIGNAL, false );
job_ad->Assign( ATTR_NUM_CKPTS, 0 );
job_ad->Assign( ATTR_NUM_JOB_STARTS, 0 );
job_ad->Assign( ATTR_NUM_RESTARTS, 0 );
job_ad->Assign( ATTR_NUM_SYSTEM_HOLDS, 0 );
job_ad->Assign( ATTR_JOB_COMMITTED_TIME, 0 );
job_ad->Assign( ATTR_CUMULATIVE_SLOT_TIME, 0 );
job_ad->Assign( ATTR_COMMITTED_SLOT_TIME, 0 );
job_ad->Assign( ATTR_TOTAL_SUSPENSIONS, 0 );
job_ad->Assign( ATTR_LAST_SUSPENSION_TIME, 0 );
job_ad->Assign( ATTR_CUMULATIVE_SUSPENSION_TIME, 0 );
job_ad->Assign( ATTR_COMMITTED_SUSPENSION_TIME, 0 );
job_ad->Assign( ATTR_JOB_ROOT_DIR, "/" );
job_ad->Assign( ATTR_MIN_HOSTS, 1 );
job_ad->Assign( ATTR_MAX_HOSTS, 1 );
job_ad->Assign( ATTR_CURRENT_HOSTS, 0 );
job_ad->Assign( ATTR_WANT_REMOTE_SYSCALLS, false );
job_ad->Assign( ATTR_WANT_CHECKPOINT, false );
job_ad->Assign( ATTR_WANT_REMOTE_IO, true );
job_ad->Assign( ATTR_JOB_STATUS, IDLE );
job_ad->Assign( ATTR_ENTERED_CURRENT_STATUS, (int)time(NULL) );
job_ad->Assign( ATTR_JOB_PRIO, 0 );
job_ad->Assign( ATTR_NICE_USER, false );
job_ad->Assign( ATTR_JOB_NOTIFICATION, NOTIFY_NEVER );
job_ad->Assign( ATTR_IMAGE_SIZE, 100 );
job_ad->Assign( ATTR_JOB_IWD, "/tmp" );
job_ad->Assign( ATTR_JOB_INPUT, NULL_FILE );
job_ad->Assign( ATTR_JOB_OUTPUT, NULL_FILE );
job_ad->Assign( ATTR_JOB_ERROR, NULL_FILE );
// Not sure what to do with these. If stdin/out/err is unset in the
// submit file, condor_submit sets In/Out/Err to the NULL_FILE and
// these transfer attributes to false. Otherwise, it leaves the
// transfer attributes unset (which is treated as true). If we
// explicitly set these to false here, our caller needs to reset
// them to true if it changes In/Out/Err and wants the default
// behavior of transfering them. This will probably be a common
// oversite. Leaving them unset should be safe if our caller doesn't
// change In/Out/Err.
//job_ad->Assign( ATTR_TRANSFER_INPUT, false );
//job_ad->Assign( ATTR_TRANSFER_OUTPUT, false );
//job_ad->Assign( ATTR_TRANSFER_ERROR, false );
//job_ad->Assign( ATTR_TRANSFER_EXECUTABLE, false );
job_ad->Assign( ATTR_BUFFER_SIZE, 512*1024 );
job_ad->Assign( ATTR_BUFFER_BLOCK_SIZE, 32*1024 );
job_ad->Assign( ATTR_SHOULD_TRANSFER_FILES,
getShouldTransferFilesString( STF_YES ) );
job_ad->Assign( ATTR_WHEN_TO_TRANSFER_OUTPUT,
getFileTransferOutputString( FTO_ON_EXIT ) );
job_ad->Assign( ATTR_REQUIREMENTS, true );
job_ad->Assign( ATTR_PERIODIC_HOLD_CHECK, false );
job_ad->Assign( ATTR_PERIODIC_REMOVE_CHECK, false );
job_ad->Assign( ATTR_PERIODIC_RELEASE_CHECK, false );
job_ad->Assign( ATTR_ON_EXIT_HOLD_CHECK, false );
job_ad->Assign( ATTR_ON_EXIT_REMOVE_CHECK, true );
job_ad->Assign( ATTR_JOB_ARGUMENTS1, "" );
job_ad->Assign( ATTR_JOB_LEAVE_IN_QUEUE, false );
job_ad->AssignExpr( ATTR_REQUEST_MEMORY, "ifthenelse(MemoryUsage isnt undefined,MemoryUsage,( ImageSize + 1023 ) / 1024)" );
job_ad->AssignExpr( ATTR_REQUEST_DISK, "DiskUsage" );
job_ad->Assign( ATTR_DISK_USAGE, 1 );
job_ad->Assign( ATTR_REQUEST_CPUS, 1 );
// Without these, the starter will not automatically remap the stdout/err (look at sha 422735d9)
// and possibly won't put them in the correct directory.
job_ad->Assign( ATTR_STREAM_OUTPUT, false );
job_ad->Assign( ATTR_STREAM_ERROR, false );
job_ad->Assign( ATTR_VERSION, CondorVersion() );
job_ad->Assign( ATTR_PLATFORM, CondorPlatform() );
job_ad->Assign( ATTR_Q_DATE, time(NULL) );
return job_ad;
}
void
XferSummary::time_out(time_t now, char *hostaddr)
{
ClassAd info;
char line[128], *tmp;
char *str = NULL;
SetMyTypeName(info, "CkptServer");
SetTargetTypeName(info, "CkptFile");
sprintf(line, "%s = \"%s\"", ATTR_NAME, get_local_fqdn().Value());
info.Insert(line);
sprintf(line, "%s = \"%s\"", ATTR_MACHINE, hostaddr );
info.Insert(line);
sprintf(line, "%s = \"%s\"", ATTR_VERSION, CondorVersion() );
info.Insert(line);
sprintf(line, "%s = \"%s\"", ATTR_PLATFORM, CondorPlatform() );
info.Insert(line);
sprintf(line, "NumSends = %d", num_sends);
info.Insert(line);
sprintf(line, "BytesSent = %d", (int) bytes_sent);
info.Insert(line);
sprintf(line, "TimeSending = %d", time_sending);
info.Insert(line);
sprintf(line, "AvgSendBandwidth = %f", num_sends ?
tot_send_bandwidth / num_sends : 0.0);
info.Insert(line);
sprintf(line, "NumRecvs = %d", num_recvs);
info.Insert(line);
sprintf(line, "BytesReceived = %d", (int) bytes_recv);
info.Insert(line);
sprintf(line, "TimeReceiving = %d", time_recving);
info.Insert(line);
sprintf(line, "AvgReceiveBandwidth = %f", num_recvs ?
tot_recv_bandwidth / num_recvs : 0.0);
info.Insert(line);
/* ctime adds a newline at the end of the ascii conversion.... */
str = ctime(&start_time);
sprintf(line, "CkptServerIntervalStart = \"%s\"", str ? str : "Unknown\n");
tmp = strchr( line, '\n' );
if (tmp != NULL) {
/* delete the newline */
*tmp = '\"';
tmp++;
*tmp = '\0';
}
info.Insert(line);
/* ctime adds a newline at the end of the ascii conversion.... */
str = ctime(&now);
sprintf(line, "CkptServerIntervalEnd = \"%s\"", str ? str : "Unknown\n");
tmp = strchr( line, '\n' );
if (tmp != NULL) {
/* delete the newline */
*tmp = '\"';
tmp++;
*tmp = '\0';
}
info.Insert(line);
info.Assign("Disk", sysapi_disk_space(pwd.Value()));
// Send to collector
if ( Collectors ) {
dprintf(D_NETWORK, "Sending CkptServer ClassAd:\n");
dPrintAd(D_NETWORK, info);
Collectors->sendUpdates (UPDATE_CKPT_SRVR_AD, &info, NULL, true);
}
init();
}
void
Starter::exited(int status)
{
ClassAd *jobAd = NULL;
bool jobAdNeedsFree = true;
if (s_claim && s_claim->ad()) {
// real jobs in the startd have claims and job ads, boinc and perhaps others won't
jobAd = s_claim->ad();
jobAdNeedsFree = false;
} else {
// Dummy up an ad
int now = (int) time(0);
jobAd = new ClassAd();
SetMyTypeName(*jobAd, "Job");
SetTargetTypeName(*jobAd, "Machine");
jobAd->Assign(ATTR_CLUSTER_ID, now);
jobAd->Assign(ATTR_PROC_ID, 1);
jobAd->Assign(ATTR_OWNER, "boinc");
jobAd->Assign(ATTR_Q_DATE, (int)s_birthdate);
jobAd->Assign(ATTR_JOB_PRIO, 0);
jobAd->Assign(ATTR_IMAGE_SIZE, 0);
jobAd->Assign(ATTR_JOB_CMD, "boinc");
MyString gjid;
gjid.formatstr("%s#%d#%d#%d", get_local_hostname().Value(),
now, 1, now);
jobAd->Assign(ATTR_GLOBAL_JOB_ID, gjid);
}
// First, patch up the ad a little bit
jobAd->Assign(ATTR_COMPLETION_DATE, (int)time(0));
int runtime = time(0) - s_birthdate;
jobAd->Assign(ATTR_JOB_REMOTE_WALL_CLOCK, runtime);
int jobStatus = COMPLETED;
if (WIFSIGNALED(status)) {
jobStatus = REMOVED;
}
jobAd->Assign(ATTR_STARTER_EXIT_STATUS, status);
jobAd->Assign(ATTR_JOB_STATUS, jobStatus);
if (s_claim) {
bool badputFromDraining = s_claim->getBadputCausedByDraining();
jobAd->Assign(ATTR_BADPUT_CAUSED_BY_DRAINING, badputFromDraining);
bool badputFromPreemption = s_claim->getBadputCausedByPreemption();
jobAd->Assign(ATTR_BADPUT_CAUSED_BY_PREEMPTION, badputFromPreemption);
}
AppendHistory(jobAd);
WritePerJobHistoryFile(jobAd, true /* use gjid for filename*/);
if (jobAdNeedsFree) {
delete jobAd;
}
// Make sure our time isn't going to go off.
cancelKillTimer();
// Just for good measure, try to kill what's left of our whole
// pid family.
if (daemonCore->Kill_Family(s_pid) == FALSE) {
dprintf(D_ALWAYS,
"error killing process family of starter with pid %u\n",
s_pid);
}
// Now, delete any files lying around.
ASSERT( executeDir() );
if ( encryptedExecuteDir() ) {
// Remove $EXECUTE/encryptedX/dir_pid/*
cleanup_execute_dir( s_pid, encryptedExecuteDir(), true );
s_encrypted_execute_dir = "";
} else {
// Remove $EXECUTE/dir_pid/*
cleanup_execute_dir( s_pid, executeDir() );
}
#if defined(LINUX)
if( param_boolean( "GLEXEC_STARTER", false ) ) {
cleanupAfterGlexec();
}
#endif
}
/*! handle a log Entry: work with a job queue collection.
* (not with DBMS directry)
*/
QuillErrCode
JobQueueDBManager::processLogEntry(int op_type, JobQueueCollection* jobQueue)
{
char *key, *mytype, *targettype, *name, *value;
key = mytype = targettype = name = value = NULL;
QuillErrCode st = QUILL_SUCCESS;
int job_id_type;
char cid[512];
char pid[512];
// REMEMBER:
// each get*ClassAdBody() funtion allocates the memory of
// parameters. Therefore, they all must be deallocated here,
// and they are at the end of the routine
switch(op_type) {
case CondorLogOp_NewClassAd: {
if (caLogParser->getNewClassAdBody(key, mytype, targettype) == QUILL_FAILURE)
{
st = QUILL_FAILURE;
break;
}
job_id_type = getProcClusterIds(key, cid, pid);
ClassAd* ad = new ClassAd();
SetMyTypeName(*ad, "Job");
SetTargetTypeName(*ad, "Machine");
switch(job_id_type) {
case IS_CLUSTER_ID:
jobQueue->insertClusterAd(cid, ad);
break;
case IS_PROC_ID:
jobQueue->insertProcAd(cid, pid, ad);
break;
default:
dprintf(D_ALWAYS, "[QUILL++] New ClassAd --- ERROR\n");
st = QUILL_FAILURE;
break;
}
break;
}
case CondorLogOp_DestroyClassAd: {
if (caLogParser->getDestroyClassAdBody(key) == QUILL_FAILURE) {
st = QUILL_FAILURE;
break;
}
job_id_type = getProcClusterIds(key, cid, pid);
switch(job_id_type) {
case IS_CLUSTER_ID:
jobQueue->removeClusterAd(cid);
break;
case IS_PROC_ID: {
ClassAd *clusterad = jobQueue->find(cid);
ClassAd *procad = jobQueue->find(cid,pid);
if(!clusterad || !procad) {
dprintf(D_ALWAYS,
"[QUILL++] Destroy ClassAd --- Cannot find clusterad "
"or procad in memory job queue");
st = QUILL_FAILURE;
break;
}
jobQueue->removeProcAd(cid, pid);
break;
}
default:
dprintf(D_ALWAYS, "[QUILL++] Destroy ClassAd --- ERROR\n");
st = QUILL_FAILURE;
break;
}
break;
}
case CondorLogOp_SetAttribute: {
if (caLogParser->getSetAttributeBody(key, name, value) == QUILL_FAILURE) {
st = QUILL_FAILURE;
break;
}
job_id_type = getProcClusterIds(key, cid, pid);
char tmp[512];
snprintf(tmp, 512, "%s = %s", name, value);
switch (job_id_type) {
case IS_CLUSTER_ID: {
ClassAd* ad = jobQueue->findClusterAd(cid);
if (ad != NULL) {
ad->Insert(tmp);
}
else {
dprintf(D_ALWAYS, "[QUILL++] ERROR: There is no such Cluster Ad[%s]\n", cid);
}
break;
}
case IS_PROC_ID: {
ClassAd* ad = jobQueue->findProcAd(cid, pid);
if (ad != NULL) {
ad->Insert(tmp);
}
else {
dprintf(D_ALWAYS, "[QUILL++] ERROR: There is no such Proc Ad[%s.%s]\n", cid, pid);
}
break;
}
default:
dprintf(D_ALWAYS, "[QUILL++] Set Attribute --- ERROR\n");
st = QUILL_FAILURE;
break;
}
break;
}
case CondorLogOp_DeleteAttribute: {
if (caLogParser->getDeleteAttributeBody(key, name) == QUILL_FAILURE) {
st = QUILL_FAILURE;
break;
}
job_id_type = getProcClusterIds(key, cid, pid);
switch(job_id_type) {
case IS_CLUSTER_ID: {
ClassAd* ad = jobQueue->findClusterAd(cid);
ad->Delete(name);
break;
}
case IS_PROC_ID: {
ClassAd* ad = jobQueue->findProcAd(cid, pid);
ad->Delete(name);
break;
}
default:
dprintf(D_ALWAYS, "[QUILL++] Delete Attribute --- ERROR\n");
st = QUILL_FAILURE;
}
break;
}
case CondorLogOp_BeginTransaction:
st = processBeginTransaction(true);
break;
case CondorLogOp_EndTransaction:
st = processEndTransaction(true);
break;
// skip the log historical sequence number command
case CondorLogOp_LogHistoricalSequenceNumber:
break;
default:
printf("[QUILL++] Unsupported Job Queue Command\n");
st = QUILL_FAILURE;
break;
}
// pointers are release
if (key != NULL) free(key);
if (mytype != NULL) free(mytype);
if (targettype != NULL) free(targettype);
if (name != NULL) free(name);
if (value != NULL) free(value);
return st;
}
ClassAd *CollectorEngine::
collect (int command,ClassAd *clientAd,const condor_sockaddr& from,int &insert,Sock *sock)
{
ClassAd *retVal;
ClassAd *pvtAd;
int insPvt;
AdNameHashKey hk;
HashString hashString;
static int repeatStartdAds = -1; // for debugging
ClassAd *clientAdToRepeat = NULL;
_condor_auto_accum_runtime<collector_runtime_probe> rt(CollectorEngine_rucc_runtime);
double rt_last = rt.begin;
if (repeatStartdAds == -1) {
repeatStartdAds = param_integer("COLLECTOR_REPEAT_STARTD_ADS",0);
}
if( !ValidateClassAd(command,clientAd,sock) ) {
return NULL;
}
CollectorEngine_rucc_validateAd_runtime.Add(rt.tick(rt_last));
// mux on command
switch (command)
{
case UPDATE_STARTD_AD:
case UPDATE_STARTD_AD_WITH_ACK:
if ( repeatStartdAds > 0 ) {
clientAdToRepeat = new ClassAd(*clientAd);
}
if (!makeStartdAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
CollectorEngine_rucc_makeHashKey_runtime.Add(rt.tick(rt_last));
retVal=updateClassAd (StartdAds, "StartdAd ", "Start",
clientAd, hk, hashString, insert, from );
if (last_updateClassAd_was_insert) { CollectorEngine_rucc_insertAd_runtime.Add(rt.tick(rt_last));
} else { CollectorEngine_rucc_updateAd_runtime.Add(rt.tick(rt_last)); }
// if we want to store private ads
if (!sock)
{
dprintf (D_ALWAYS, "Want private ads, but no socket given!\n");
break;
}
else
{
if (!(pvtAd = new ClassAd))
{
EXCEPT ("Memory error!");
}
if( !getClassAd(sock, *pvtAd) )
{
dprintf(D_FULLDEBUG,"\t(Could not get startd's private ad)\n");
delete pvtAd;
break;
}
// Fix up some stuff in the private ad that we depend on.
// We started doing this in 7.2.0, so once we no longer
// care about compatibility with stuff from before then,
// the startd could stop bothering to send these attributes.
// Queries of private ads depend on the following:
SetMyTypeName( *pvtAd, STARTD_ADTYPE );
// Negotiator matches up private ad with public ad by
// using the following.
if( retVal ) {
CopyAttribute( ATTR_MY_ADDRESS, *pvtAd, *retVal );
CopyAttribute( ATTR_NAME, *pvtAd, *retVal );
}
CollectorEngine_rucc_getPvtAd_runtime.Add(rt.tick(rt_last));
// insert the private ad into its hashtable --- use the same
// hash key as the public ad
(void) updateClassAd (StartdPrivateAds, "StartdPvtAd ",
"StartdPvt", pvtAd, hk, hashString, insPvt,
from );
if (last_updateClassAd_was_insert) { CollectorEngine_rucc_insertPvtAd_runtime.Add(rt.tick(rt_last));
} else { CollectorEngine_rucc_updatePvtAd_runtime.Add(rt.tick(rt_last)); }
}
// create fake duplicates of this ad, each with a different name, if
// we are told to do so. this feature exists for developer
// scalability testing.
if ( repeatStartdAds > 0 && clientAdToRepeat ) {
ClassAd *fakeAd;
int n;
char newname[150],oldname[130];
oldname[0] = '\0';
clientAdToRepeat->LookupString("Name",oldname,sizeof(oldname));
for (n=0;n<repeatStartdAds;n++) {
fakeAd = new ClassAd(*clientAdToRepeat);
snprintf(newname,sizeof(newname),
"Name=\"fake%d-%s\"",n,oldname);
fakeAd->Insert(newname);
makeStartdAdHashKey (hk, fakeAd);
hashString.Build( hk );
if (! updateClassAd (StartdAds, "StartdAd ", "Start",
fakeAd, hk, hashString, insert, from ) )
{
// don't leak memory if there is some failure
delete fakeAd;
}
}
delete clientAdToRepeat;
clientAdToRepeat = NULL;
CollectorEngine_rucc_repeatAd_runtime.Add(rt.tick(rt_last));
}
break;
case MERGE_STARTD_AD:
if (!makeStartdAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=mergeClassAd (StartdAds, "StartdAd ", "Start",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_SCHEDD_AD:
if (!makeScheddAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (ScheddAds, "ScheddAd ", "Schedd",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_SUBMITTOR_AD:
// use the same hashkey function as a schedd ad
if (!makeScheddAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
// since submittor ads always follow a schedd ad, and a master check is
// performed for schedd ads, we don't need a master check in here
hashString.Build( hk );
retVal=updateClassAd (SubmittorAds, "SubmittorAd ", "Submittor",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_LICENSE_AD:
// use the same hashkey function as a schedd ad
if (!makeLicenseAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
// since submittor ads always follow a schedd ad, and a master check is
// performed for schedd ads, we don't need a master check in here
hashString.Build( hk );
retVal=updateClassAd (LicenseAds, "LicenseAd ", "License",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_MASTER_AD:
if (!makeMasterAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (MasterAds, "MasterAd ", "Master",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_CKPT_SRVR_AD:
if (!makeCkptSrvrAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (CkptServerAds, "CkptSrvrAd ", "CkptSrvr",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_COLLECTOR_AD:
if (!makeCollectorAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (CollectorAds, "CollectorAd ", "Collector",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_STORAGE_AD:
if (!makeStorageAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (StorageAds, "StorageAd ", "Storage",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_ACCOUNTING_AD:
if (!makeAccountingAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (AccountingAds, "AccountingAd ", "Accouting",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_NEGOTIATOR_AD:
if (!makeNegotiatorAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
if (m_allowOnlyOneNegotiator) {
// first, purge all the existing negotiator ads, since we
// want to enforce that *ONLY* 1 negotiator is in the
// collector any given time.
purgeHashTable( NegotiatorAds );
}
retVal=updateClassAd (NegotiatorAds, "NegotiatorAd ", "Negotiator",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_HAD_AD:
if (!makeHadAdHashKey (hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (HadAds, "HadAd ", "HAD",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_GRID_AD:
if (!makeGridAdHashKey(hk, clientAd))
{
dprintf (D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build( hk );
retVal=updateClassAd (GridAds, "GridAd ", "Grid",
clientAd, hk, hashString, insert, from );
break;
case UPDATE_AD_GENERIC:
{
const char *type_str = GetMyTypeName(*clientAd);
if (type_str == NULL) {
dprintf(D_ALWAYS, "collect: UPDATE_AD_GENERIC: ad has no type\n");
insert = -3;
retVal = 0;
break;
}
MyString type(type_str);
CollectorHashTable *cht = findOrCreateTable(type);
if (cht == NULL) {
dprintf(D_ALWAYS, "collect: findOrCreateTable failed\n");
insert = -3;
retVal = 0;
break;
}
if (!makeGenericAdHashKey (hk, clientAd))
{
dprintf(D_ALWAYS, "Could not make hashkey --- ignoring ad\n");
insert = -3;
retVal = 0;
break;
}
hashString.Build(hk);
retVal = updateClassAd(*cht, type_str, type_str, clientAd,
hk, hashString, insert, from);
break;
}
case QUERY_STARTD_ADS:
case QUERY_SCHEDD_ADS:
case QUERY_MASTER_ADS:
case QUERY_SUBMITTOR_ADS:
case QUERY_CKPT_SRVR_ADS:
case QUERY_STARTD_PVT_ADS:
case QUERY_COLLECTOR_ADS:
case QUERY_NEGOTIATOR_ADS:
case QUERY_HAD_ADS:
case QUERY_GENERIC_ADS:
case INVALIDATE_STARTD_ADS:
case INVALIDATE_SCHEDD_ADS:
case INVALIDATE_MASTER_ADS:
case INVALIDATE_CKPT_SRVR_ADS:
case INVALIDATE_SUBMITTOR_ADS:
case INVALIDATE_COLLECTOR_ADS:
case INVALIDATE_NEGOTIATOR_ADS:
case INVALIDATE_HAD_ADS:
case INVALIDATE_ADS_GENERIC:
// these are not implemented in the engine, but we allow another
// daemon to detect that these commands have been given
insert = -2;
retVal = 0;
break;
default:
dprintf (D_ALWAYS, "Received illegal command: %d\n", command);
insert = -1;
retVal = 0;
}
if (command != UPDATE_STARTD_AD && command != UPDATE_STARTD_AD_WITH_ACK) {
CollectorEngine_rucc_other_runtime.Add(rt.tick(rt_last));
}
// return the updated ad
return retVal;
}
