// wait for the transitioner to create a result for the given WU.
// This keeps us from getting infinitely far ahead of the transitioner
// (e.g. if the transitioner isn't running)
//
void wait_for_results(int wu_id) {
DB_RESULT result;
int count, retval;
char buf[256];
sprintf(buf, "where workunitid=%d", wu_id);
while (1) {
retval = result.count(count, buf);
log_messages.printf(MSG_DEBUG, "result.count for %d returned %d, error: %s\n",
wu_id, count, boincerror(retval)
);
if (retval) {
log_messages.printf(MSG_CRITICAL, "result.count: %s\n", boincerror(retval));
exit(1);
}
if (count > 0) return;
sleep(10);
check_stop_daemons();
}
}
int count_results(char* query, int& n) {
DB_RESULT result;
return result.count(n, query);
}
// Do checks that require DB access for whether we can send this job,
// and return:
// 0 if OK to send
// 1 if can't send to this host
// 2 if can't send to ANY host
//
int slow_check(
WU_RESULT& wu_result, // the job cache entry.
// We may refresh its hr_class and app_version_id fields.
APP* app,
BEST_APP_VERSION* bavp // the app version to be used
) {
int n, retval;
DB_RESULT result;
char buf[256];
WORKUNIT& wu = wu_result.workunit;
// Don't send if we've already sent a result of this WU to this user.
//
if (config.one_result_per_user_per_wu) {
sprintf(buf,
"where workunitid=%d and userid=%d", wu.id, g_reply->user.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%s)\n", boincerror(retval)
);
return 1;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [USER#%d] already has %d result(s) for [WU#%u]\n",
g_reply->user.id, n, wu.id
);
}
return 1;
}
}
} else if (config.one_result_per_host_per_wu) {
// Don't send if we've already sent a result of this WU to this host.
// We only have to check this if we don't send one result per user.
//
sprintf(buf,
"where workunitid=%d and hostid=%d", wu.id, g_reply->host.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%s)\n", boincerror(retval)
);
return 1;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] already has %d result(s) for [WU#%u]\n",
g_reply->host.id, n, wu.id
);
}
return 1;
}
}
}
// Checks that require looking up the WU.
// Lump these together so we only do 1 lookup
//
if (app_hr_type(*app) || app->homogeneous_app_version) {
DB_WORKUNIT db_wu;
db_wu.id = wu.id;
int vals[3];
retval = db_wu.get_field_ints(
"hr_class, app_version_id, error_mask", 3, vals
);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"can't get fields for [WU#%u]: %s\n", db_wu.id, boincerror(retval)
);
return 1;
}
// check wu.error_mask
//
if (vals[2] != 0) {
return 2;
}
if (app_hr_type(*app)) {
wu.hr_class = vals[0];
if (already_sent_to_different_hr_class(wu, *app)) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] [WU#%u %s] is assigned to different HR class\n",
g_reply->host.id, wu.id, wu.name
);
}
// Mark the workunit as infeasible.
// This ensures that jobs already assigned to an HR class
// are processed first.
//
wu_result.infeasible_count++;
return 1;
}
}
if (app->homogeneous_app_version) {
int wu_avid = vals[1];
wu.app_version_id = wu_avid;
if (wu_avid && wu_avid != bavp->avp->id) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] [WU#%u %s] is assigned to different app version\n",
g_reply->host.id, wu.id, wu.name
);
}
wu_result.infeasible_count++;
return 1;
}
}
}
return 0;
}
// do slow checks (ones that require DB access)
//
static bool slow_check(WU_RESULT& wu_result, WORKUNIT& wu, APP* app) {
int n, retval;
DB_RESULT result;
char buf[256];
// Don't send if we've already sent a result of this WU to this user.
//
if (config.one_result_per_user_per_wu) {
sprintf(buf,
"where workunitid=%d and userid=%d",
wu_result.workunit.id, g_reply->user.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%d)\n", retval
);
return false;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [USER#%d] already has %d result(s) for [WU#%d]\n",
g_reply->user.id, n, wu_result.workunit.id
);
}
return false;
}
}
} else if (config.one_result_per_host_per_wu) {
// Don't send if we've already sent a result
// of this WU to this host.
// We only have to check this
// if we don't send one result per user.
//
sprintf(buf,
"where workunitid=%d and hostid=%d",
wu_result.workunit.id, g_reply->host.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%d)\n", retval
);
return false;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] already has %d result(s) for [WU#%d]\n",
g_reply->host.id, n, wu_result.workunit.id
);
}
return false;
}
}
}
if (app_hr_type(*app)) {
if (already_sent_to_different_platform_careful(
wu_result.workunit, *app
)) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] [WU#%d %s] is assigned to different platform\n",
g_reply->host.id, wu.id, wu.name
);
}
// Mark the workunit as infeasible.
// This ensures that jobs already assigned to a platform
// are processed first.
//
wu_result.infeasible_count++;
return false;
}
}
return true;
}
int count_results(char* query, int& n) {
DB_RESULT result;
int retval = result.count(n, query);
if (retval) return retval;
return 0;
}
bool wu_is_infeasible_slow(
WU_RESULT& wu_result, SCHEDULER_REQUEST& sreq, SCHEDULER_REPLY& reply
) {
char buf[256];
int retval;
int n;
DB_RESULT result;
// Don't send if we've already sent a result of this WU to this user.
//
if (config.one_result_per_user_per_wu) {
sprintf(buf,
"where workunitid=%d and userid=%d",
wu_result.workunit.id, g_reply->user.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%d)\n", retval
);
return true;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] send_work: user %d already has %d result(s) for WU %d\n",
g_reply->user.id, n, wu_result.workunit.id
);
}
return true;
}
}
} else if (config.one_result_per_host_per_wu) {
// Don't send if we've already sent a result
// of this WU to this host.
// We only have to check this
// if we don't send one result per user.
//
sprintf(buf,
"where workunitid=%d and hostid=%d",
wu_result.workunit.id, g_reply->host.id
);
retval = result.count(n, buf);
if (retval) {
log_messages.printf(MSG_CRITICAL,
"send_work: can't get result count (%d)\n", retval
);
return true;
} else {
if (n>0) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] send_work: host %d already has %d result(s) for WU %d\n",
g_reply->host.id, n, wu_result.workunit.id
);
}
return true;
}
}
}
APP* app = ssp->lookup_app(wu_result.workunit.appid);
WORKUNIT wu = wu_result.workunit;
if (app_hr_type(*app)) {
if (already_sent_to_different_platform_careful(wu, *app)) {
if (config.debug_send) {
log_messages.printf(MSG_NORMAL,
"[send] [HOST#%d] [WU#%d %s] WU is infeasible (assigned to different platform)\n",
g_reply->host.id, wu.id, wu.name
);
}
// Mark the workunit as infeasible.
// This ensures that jobs already assigned to a platform
// are processed first.
//
wu_result.infeasible_count++;
return true;
}
}
return false;
}
