int main(int argc, char** argv) {
HR_INFO hri;
int retval;
for (int i=0; i<argc; i++) {
if (!strcmp(argv[i], "--help") || !strcmp(argv[i], "-h")) {
usage(argv);
}
}
check_stop_daemons();
retval = config.parse_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't parse config.xml: %s\n", boincerror(retval)
);
exit(1);
}
retval = boinc_db.open(
config.db_name, config.db_host, config.db_user, config.db_passwd
);
if (retval) {
log_messages.printf(MSG_CRITICAL, "Can't open DB\n");
exit(1);
}
log_messages.printf(MSG_NORMAL, "Starting\n");
boinc_db.set_isolation_level(READ_UNCOMMITTED);
hri.init();
hri.scan_db();
hri.write_file();
hri.perf_info.write_file();
log_messages.printf(MSG_NORMAL, "Finished\n");
}
// write a summary of feeder state to stderr
//
void show_state(int) {
ssp->show(stderr);
if (config.hr_allocate_slots) {
hr_info.show(stderr);
}
}
// see if we're using HR, and if so initialize the necessary data structures
//
void hr_init() {
int i, retval;
bool apps_differ = false;
bool some_app_uses_hr = false;
int hrt, hr_type0 = ssp->apps[0].homogeneous_redundancy;
using_hr = false;
for (i=0; i<ssp->napps; i++) {
hrt = ssp->apps[i].homogeneous_redundancy;
if (hrt <0 || hrt >= HR_NTYPES) {
log_messages.printf(MSG_CRITICAL,
"HR type %d out of range for app %d\n", hrt, i
);
exit(1);
}
if (hrt) some_app_uses_hr = true;
if (hrt != hr_type0) apps_differ = true;
}
if (config.homogeneous_redundancy) {
log_messages.printf(MSG_NORMAL,
"config HR is %d\n", config.homogeneous_redundancy
);
hrt = config.homogeneous_redundancy;
if (hrt < 0 || hrt >= HR_NTYPES) {
log_messages.printf(MSG_CRITICAL,
"Main HR type %d out of range\n", hrt
);
exit(1);
}
if (some_app_uses_hr) {
log_messages.printf(MSG_CRITICAL,
"You can specify HR at global or app level, but not both\n"
);
exit(1);
}
for (i=0; i<ssp->napps; i++) {
ssp->apps[i].homogeneous_redundancy = config.homogeneous_redundancy;
ssp->apps[i].weight = 1;
}
} else {
if (some_app_uses_hr) {
if (apps_differ && !all_apps) {
log_messages.printf(MSG_CRITICAL,
"You must use --allapps if apps have different HR\n"
);
exit(1);
}
} else {
return; // HR not being used
}
}
using_hr = true;
if (config.hr_allocate_slots) {
hr_info.init();
retval = hr_info.read_file();
if (retval) {
log_messages.printf(MSG_CRITICAL,
"Can't read HR info file: %s\n", boincerror(retval)
);
exit(1);
}
// find the weight for each HR type
//
for (i=0; i<ssp->napps; i++) {
hrt = ssp->apps[i].homogeneous_redundancy;
hr_info.type_weights[hrt] += ssp->apps[i].weight;
hr_info.type_being_used[hrt] = true;
}
// compute the slot allocations for HR classes
//
hr_info.allocate(ssp->max_wu_results);
hr_info.show(stderr);
}
}
// Enumerate jobs from DB until find one that is not already in the work array.
// If find one, return true.
// If reach end of enum for second time on this array scan, return false
//
static bool get_job_from_db(
DB_WORK_ITEM& wi, // enumerator to get job from
int app_index, // if using --allapps, the app index
int& enum_phase,
int& ncollisions
) {
bool collision;
int retval, j, enum_size;
char select_clause[256];
if (all_apps) {
sprintf(select_clause, "%s and r1.appid=%lu",
mod_select_clause, ssp->apps[app_index].id
);
enum_size = enum_sizes[app_index];
} else {
safe_strcpy(select_clause, mod_select_clause);
enum_size = enum_limit;
}
int hrt = ssp->apps[app_index].homogeneous_redundancy;
while (1) {
if (hrt && config.hr_allocate_slots) {
retval = wi.enumerate_all(enum_size, select_clause);
} else {
retval = wi.enumerate(enum_size, select_clause, order_clause);
}
if (retval) {
if (retval != ERR_DB_NOT_FOUND) {
// If DB server dies, exit;
// so /start (run from crontab) will restart us eventually.
//
log_messages.printf(MSG_CRITICAL,
"DB connection lost, exiting\n"
);
exit(0);
}
// we've reach the end of the result set
//
switch (enum_phase) {
case ENUM_FIRST_PASS:
enum_phase = ENUM_SECOND_PASS;
ncollisions = 0;
// disregard collisions - maybe we'll find new jobs
break;
case ENUM_SECOND_PASS:
enum_phase = ENUM_OVER;
return false;
}
log_messages.printf(MSG_NORMAL,
"restarted enumeration for appid %lu\n",
ssp->apps[app_index].id
);
} else {
// Check for invalid application ID
//
if (!ssp->lookup_app(wi.wu.appid)) {
#if 0
log_messages.printf(MSG_CRITICAL,
"result [RESULT#%u] has bad appid %d; clean up your DB!\n",
wi.res_id, wi.wu.appid
);
#endif
continue;
}
// if the WU had an error, mark result as DIDNT_NEED
//
if (wi.wu.error_mask) {
char buf[256];
DB_RESULT result;
result.id = wi.res_id;
sprintf(buf, "server_state=%d, outcome=%d",
RESULT_SERVER_STATE_OVER,
RESULT_OUTCOME_DIDNT_NEED
);
result.update_field(buf);
log_messages.printf(MSG_NORMAL,
"[RESULT#%lu] WU had error, marking as DIDNT_NEED\n",
wi.res_id
);
continue;
}
// Check for collision (i.e. this result already is in the array)
//
collision = false;
for (j=0; j<ssp->max_wu_results; j++) {
if (ssp->wu_results[j].state != WR_STATE_EMPTY && ssp->wu_results[j].resultid == wi.res_id) {
// If the result is already in shared mem,
// and another instance of the WU has been sent,
// bump the infeasible count to encourage
// it to get sent more quickly
//
if (ssp->wu_results[j].infeasible_count == 0) {
if (wi.wu.hr_class > 0) {
ssp->wu_results[j].infeasible_count++;
}
}
ncollisions++;
collision = true;
log_messages.printf(MSG_DEBUG,
"result [RESULT#%lu] already in array\n", wi.res_id
);
break;
}
}
if (collision) {
continue;
}
// if using HR, check whether we've exceeded quota for this class
//
if (hrt && config.hr_allocate_slots) {
if (!hr_info.accept(hrt, wi.wu.hr_class)) {
log_messages.printf(MSG_DEBUG,
"rejecting [RESULT#%lu] because HR class %d/%d over quota\n",
wi.res_id, hrt, wi.wu.hr_class
);
continue;
}
}
return true;
}
}
return false; // never reached
}
// write a summary of feeder state to stderr
//
void show_state(int) {
ssp->show(stderr);
hr_info.show(stderr);
}