static void catalog_update_udp( const char *host, const char *address, int port, const char *text )
{
debug(D_DEBUG, "sending update via udp to %s(%s):%d", host, address, port);
struct datagram *d = datagram_create(DATAGRAM_PORT_ANY);
if(!d) return;
datagram_send(d, text, strlen(text), address, port);
datagram_delete(d);
}
int advertise_master_to_catalog(const char *catalog_host, int catalog_port, const char *project_name, const char *master_address, struct work_queue_stats *s, struct work_queue_resources *r, const char *workers_by_pool ) {
char address[DATAGRAM_ADDRESS_MAX];
char owner[USERNAME_MAX];
buffer_t *buffer = NULL;
const char *text;
size_t text_size;
if(!outgoing_datagram) {
outgoing_datagram = datagram_create(0);
if(!outgoing_datagram) {
fprintf(stderr, "Failed to advertise master to catalog server: couldn't create outgoing udp datagram!\n");
return 0;
}
}
if(!username_get(owner)) {
strcpy(owner,"unknown");
}
buffer = buffer_create();
int total_workers_working = s->workers_busy + s->workers_full;
int total_workers = total_workers_working + s->workers_ready;
debug(D_WQ,"%s advertising resources to the Catalog -- cores:%d memory:%d disk:%d\n",project_name,r->cores.total,r->memory.total,r->disk.total); //debug to see if information is being passed
buffer_printf(buffer,
"type wq_master\n"
"project %s\nstarttime %llu\npriority %d\n"
"port %d\nlifetime %d\n"
"tasks_waiting %d\ntasks_complete %d\ntasks_running %d\ntotal_tasks_dispatched %d\n"
"workers_init %d\nworkers_ready %d\nworkers_busy %d\nworkers %d\nworkers_by_pool %s\n"
"cores_total %d\nmemory_total %d\ndisk_total %d\n"
"capacity %d\n"
"my_master %s\n"
"version %d.%d.%s\nowner %s",
project_name, (s->start_time)/1000000, s->priority,
s->port, WORK_QUEUE_CATALOG_MASTER_AD_LIFETIME,
s->tasks_waiting, s->total_tasks_complete, s->tasks_running, s->total_tasks_dispatched,
s->workers_init, s->workers_ready, total_workers_working, total_workers, workers_by_pool,
r->cores.total, r->memory.total, r->disk.total,
s->capacity,
master_address,
CCTOOLS_VERSION_MAJOR, CCTOOLS_VERSION_MINOR, CCTOOLS_VERSION_MICRO, owner);
text = buffer_tostring(buffer, &text_size);
if(domain_name_cache_lookup(catalog_host, address)) {
debug(D_WQ, "Advertising master status to the catalog server at %s:%d ...", catalog_host, catalog_port);
datagram_send(outgoing_datagram, text, strlen(text), address, catalog_port);
}
buffer_delete(buffer);
return 1;
}
int advertise_pool_decision_to_catalog(const char *catalog_host, int catalog_port, const char *pool_name, pid_t pid, time_t start_time, const char *decision, int workers_requested)
{
char address[DATAGRAM_ADDRESS_MAX];
char owner[USERNAME_MAX];
buffer_t B;
const char *text;
size_t text_size;
static time_t last_update_time = 0;
if(time(0) - last_update_time < WORK_QUEUE_CATALOG_POOL_UPDATE_INTERVAL) return 1;
if(!outgoing_datagram) {
outgoing_datagram = datagram_create(0);
if(!outgoing_datagram) {
fprintf(stderr, "Couldn't create outgoing udp port, thus work queue master info won't be sent to the catalog server!\n");
return 0;
}
}
if(!username_get(owner)) {
strcpy(owner,"unknown");
}
// port = MAX_TCP_PORT + process id, this is for the catalog server to
// distinguish the worker pools from the same host. See make_hash_key()
// function in catalog_server.c
INT64_T port = 65535 + pid;
buffer_init(&B);
buffer_abortonfailure(&B, 1);
buffer_printf(&B, "type wq_pool\npool_name %s\nport %" PRId64 "\nstarttime %llu\ndecision %s\nworkers_requested %d\nowner %s\nlifetime %d", pool_name, port, (unsigned long long) start_time, decision, workers_requested, owner, WORK_QUEUE_CATALOG_POOL_AD_LIFETIME);
text = buffer_tostring(&B, &text_size);
debug(D_WQ, "Pool AD: \n%s\n", text);
if(domain_name_cache_lookup(catalog_host, address)) {
debug(D_WQ, "Sending the pool decision to the catalog server at %s:%d ...", catalog_host, catalog_port);
datagram_send(outgoing_datagram, text, text_size, address, catalog_port);
}
buffer_free(&B);
last_update_time = time(0);
return 1;
}
int advertise_master_to_catalog(const char *catalog_host, int catalog_port, const char *project_name, struct work_queue_stats *s, const char *workers_by_pool, int now) {
char address[DATAGRAM_ADDRESS_MAX];
char owner[USERNAME_MAX];
buffer_t *buffer = NULL;
const char *text;
size_t text_size;
static time_t last_update_time = 0;
if(!now) {
if(time(0) - last_update_time < WORK_QUEUE_CATALOG_UPDATE_INTERVAL) return 1;
}
if(!outgoing_datagram) {
outgoing_datagram = datagram_create(0);
if(!outgoing_datagram) {
fprintf(stderr, "Failed to advertise master to catalog server: couldn't create outgoing udp datagram!\n");
return 0;
}
}
if(!username_get(owner)) {
strcpy(owner,"unknown");
}
buffer = buffer_create();
buffer_printf(buffer, "type wq_master\nproject %s\nstart_time %llu\npriority %d\nport %d\nlifetime %d\ntasks_waiting %d\ntasks_complete %d\ntask_running %d\ntotal_tasks_dispatched %d\nworkers_init %d\nworkers_ready %d\nworkers_busy %d\nworkers %d\nworkers_by_pool %s\ncapacity %d\nversion %d.%d.%d\nowner %s", project_name, s->start_time, s->priority, s->port, WORK_QUEUE_CATALOG_LIFETIME, s->tasks_waiting, s->total_tasks_complete, s->workers_busy, s->total_tasks_dispatched, s->workers_init, s->workers_ready, s->workers_busy, s->workers_ready + s->workers_busy, workers_by_pool, s->capacity, CCTOOLS_VERSION_MAJOR, CCTOOLS_VERSION_MINOR, CCTOOLS_VERSION_MICRO, owner);
text = buffer_tostring(buffer, &text_size);
if(domain_name_cache_lookup(catalog_host, address)) {
debug(D_WQ, "Advertising master status to the catalog server at %s:%d ...", catalog_host, catalog_port);
datagram_send(outgoing_datagram, text, strlen(text), address, catalog_port);
}
buffer_delete(buffer);
last_update_time = time(0);
return 1;
}
int advertise_pool_decision_to_catalog(const char *catalog_host, int catalog_port, const char *pool_name, const char *decision)
{
char address[DATAGRAM_ADDRESS_MAX];
char owner[USERNAME_MAX];
buffer_t *buffer = NULL;
const char *text;
size_t text_size;
static time_t last_update_time = 0;
if(time(0) - last_update_time < WORK_QUEUE_CATALOG_UPDATE_INTERVAL) return 1;
if(!outgoing_datagram) {
outgoing_datagram = datagram_create(0);
if(!outgoing_datagram) {
fprintf(stderr, "Couldn't create outgoing udp port, thus work queue master info won't be sent to the catalog server!\n");
return 0;
}
}
if(!username_get(owner)) {
strcpy(owner,"unknown");
}
buffer = buffer_create();
buffer_printf(buffer, "type wq_pool\npool_name %s\ndecision %s\nowner %s", pool_name, decision, owner);
text = buffer_tostring(buffer, &text_size);
debug(D_WQ, "Pool AD: \n%s\n", text);
if(domain_name_cache_lookup(catalog_host, address)) {
debug(D_WQ, "Sending the pool decision to the catalog server at %s:%d ...", catalog_host, catalog_port);
datagram_send(outgoing_datagram, text, text_size, address, catalog_port);
}
buffer_delete(buffer);
last_update_time = time(0);
return 1;
}
int main(int argc, char *argv[])
{
struct link *link, *list_port = 0;
char ch;
time_t current;
int is_daemon = 0;
char *pidfile = NULL;
outgoing_host_list = list_create();
debug_config(argv[0]);
while((ch = getopt(argc, argv, "bB:d:hH:l:L:m:M:n:o:O:p:ST:u:U:v")) != (char) -1) {
switch (ch) {
case 'b':
is_daemon = 1;
break;
case 'B':
free(pidfile);
pidfile = strdup(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'h':
default:
show_help(argv[0]);
return 1;
case 'l':
lifetime = string_time_parse(optarg);
break;
case 'L':
logfilename = strdup(optarg);
break;
case 'H':
history_dir = strdup(optarg);
break;
case 'm':
child_procs_max = atoi(optarg);
break;
case 'M':
max_server_size = string_metric_parse(optarg);
break;
case 'n':
preferred_hostname = optarg;
break;
case 'o':
free(debug_filename);
debug_filename = strdup(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'p':
port = atoi(optarg);
break;
case 'S':
fork_mode = 0;
break;
case 'T':
child_procs_timeout = string_time_parse(optarg);
break;
case 'u':
list_push_head(outgoing_host_list, xxstrdup(optarg));
break;
case 'U':
outgoing_timeout = string_time_parse(optarg);
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
}
}
if (is_daemon) daemonize(0, pidfile);
debug_config_file(debug_filename);
cctools_version_debug(D_DEBUG, argv[0]);
if(logfilename) {
logfile = fopen(logfilename,"a");
if(!logfile) fatal("couldn't open %s: %s\n",optarg,strerror(errno));
}
current = time(0);
debug(D_ALL, "*** %s starting at %s", argv[0], ctime(¤t));
if(!list_size(outgoing_host_list)) {
list_push_head(outgoing_host_list, CATALOG_HOST_DEFAULT);
}
install_handler(SIGPIPE, ignore_signal);
install_handler(SIGHUP, ignore_signal);
install_handler(SIGCHLD, ignore_signal);
install_handler(SIGINT, shutdown_clean);
install_handler(SIGTERM, shutdown_clean);
install_handler(SIGQUIT, shutdown_clean);
install_handler(SIGALRM, shutdown_clean);
if(!preferred_hostname) {
domain_name_cache_guess(hostname);
preferred_hostname = hostname;
}
username_get(owner);
starttime = time(0);
table = nvpair_database_create(history_dir);
if(!table)
fatal("couldn't create directory %s: %s\n",history_dir,strerror(errno));
update_dgram = datagram_create(port);
if(!update_dgram)
fatal("couldn't listen on udp port %d", port);
outgoing_dgram = datagram_create(0);
if(!outgoing_dgram)
fatal("couldn't create outgoing udp port");
list_port = link_serve(port);
if(!list_port)
fatal("couldn't listen on tcp port %d", port);
while(1) {
fd_set rfds;
int ufd = datagram_fd(update_dgram);
int lfd = link_fd(list_port);
int result, maxfd;
struct timeval timeout;
remove_expired_records();
if(time(0) > outgoing_alarm) {
update_all_catalogs(outgoing_dgram);
outgoing_alarm = time(0) + outgoing_timeout;
}
while(1) {
int status;
pid_t pid = waitpid(-1, &status, WNOHANG);
if(pid>0) {
child_procs_count--;
continue;
} else {
break;
}
}
FD_ZERO(&rfds);
FD_SET(ufd, &rfds);
if(child_procs_count < child_procs_max) {
FD_SET(lfd, &rfds);
}
maxfd = MAX(ufd, lfd) + 1;
timeout.tv_sec = 5;
timeout.tv_usec = 0;
result = select(maxfd, &rfds, 0, 0, &timeout);
if(result <= 0)
continue;
if(FD_ISSET(ufd, &rfds)) {
handle_updates(update_dgram);
}
if(FD_ISSET(lfd, &rfds)) {
link = link_accept(list_port, time(0) + 5);
if(link) {
if(fork_mode) {
pid_t pid = fork();
if(pid == 0) {
alarm(child_procs_timeout);
handle_query(link);
_exit(0);
} else if (pid>0) {
child_procs_count++;
}
} else {
handle_query(link);
}
link_close(link);
}
}
}
return 1;
}
int main(int argc, char *argv[])
{
struct link *link, *list_port = 0;
signed char ch;
time_t current;
int is_daemon = 0;
char *pidfile = NULL;
char *interface = NULL;
outgoing_host_list = list_create();
debug_config(argv[0]);
static const struct option long_options[] = {
{"background", no_argument, 0, 'b'},
{"pid-file", required_argument, 0, 'B'},
{"debug", required_argument, 0, 'd'},
{"help", no_argument, 0, 'h'},
{"history", required_argument, 0, 'H'},
{"lifetime", required_argument, 0, 'l'},
{"update-log", required_argument, 0, 'L'},
{"max-jobs", required_argument, 0, 'm'},
{"server-size", required_argument, 0, 'M'},
{"name", required_argument, 0, 'n'},
{"interface", required_argument, 0, 'I'},
{"debug-file", required_argument, 0, 'o'},
{"debug-rotate-max", required_argument, 0, 'O'},
{"port", required_argument, 0, 'p'},
{"single", no_argument, 0, 'S'},
{"timeout", required_argument, 0, 'T'},
{"update-host", required_argument, 0, 'u'},
{"update-interval", required_argument, 0, 'U'},
{"version", no_argument, 0, 'v'},
{"port-file", required_argument, 0, 'Z'},
{0,0,0,0}};
while((ch = getopt_long(argc, argv, "bB:d:hH:I:l:L:m:M:n:o:O:p:ST:u:U:vZ:", long_options, NULL)) > -1) {
switch (ch) {
case 'b':
is_daemon = 1;
break;
case 'B':
free(pidfile);
pidfile = strdup(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'h':
default:
show_help(argv[0]);
return 1;
case 'l':
lifetime = string_time_parse(optarg);
break;
case 'L':
logfilename = strdup(optarg);
break;
case 'H':
history_dir = strdup(optarg);
break;
case 'I':
free(interface);
interface = strdup(optarg);
break;
case 'm':
child_procs_max = atoi(optarg);
break;
case 'M':
max_server_size = string_metric_parse(optarg);
break;
case 'n':
preferred_hostname = optarg;
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'p':
port = atoi(optarg);
break;
case 'S':
fork_mode = 0;
break;
case 'T':
child_procs_timeout = string_time_parse(optarg);
break;
case 'u':
list_push_head(outgoing_host_list, xxstrdup(optarg));
break;
case 'U':
outgoing_timeout = string_time_parse(optarg);
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
case 'Z':
port_file = optarg;
port = 0;
break;
}
}
if (is_daemon) daemonize(0, pidfile);
cctools_version_debug(D_DEBUG, argv[0]);
if(logfilename) {
logfile = fopen(logfilename,"a");
if(!logfile) fatal("couldn't open %s: %s\n",optarg,strerror(errno));
}
current = time(0);
debug(D_NOTICE, "*** %s starting at %s", argv[0], ctime(¤t));
if(!list_size(outgoing_host_list)) {
list_push_head(outgoing_host_list, CATALOG_HOST_DEFAULT);
}
install_handler(SIGPIPE, ignore_signal);
install_handler(SIGHUP, ignore_signal);
install_handler(SIGCHLD, ignore_signal);
install_handler(SIGINT, shutdown_clean);
install_handler(SIGTERM, shutdown_clean);
install_handler(SIGQUIT, shutdown_clean);
install_handler(SIGALRM, shutdown_clean);
if(!preferred_hostname) {
domain_name_cache_guess(hostname);
preferred_hostname = hostname;
}
username_get(owner);
starttime = time(0);
table = nvpair_database_create(history_dir);
if(!table)
fatal("couldn't create directory %s: %s\n",history_dir,strerror(errno));
list_port = link_serve_address(interface, port);
if(list_port) {
/*
If a port was chosen automatically, read it back
so that the same one can be used for the update port.
There is the possibility that the UDP listen will
fail because that port is in use.
*/
if(port==0) {
char addr[LINK_ADDRESS_MAX];
link_address_local(list_port,addr,&port);
}
} else {
if(interface)
fatal("couldn't listen on TCP address %s port %d", interface, port);
else
fatal("couldn't listen on TCP port %d", port);
}
outgoing_dgram = datagram_create(0);
if(!outgoing_dgram)
fatal("couldn't create outgoing udp port");
update_dgram = datagram_create_address(interface, port);
if(!update_dgram) {
if(interface)
fatal("couldn't listen on UDP address %s port %d", interface, port);
else
fatal("couldn't listen on UDP port %d", port);
}
opts_write_port_file(port_file,port);
while(1) {
fd_set rfds;
int ufd = datagram_fd(update_dgram);
int lfd = link_fd(list_port);
int result, maxfd;
struct timeval timeout;
remove_expired_records();
if(time(0) > outgoing_alarm) {
update_all_catalogs(outgoing_dgram);
outgoing_alarm = time(0) + outgoing_timeout;
}
while(1) {
int status;
pid_t pid = waitpid(-1, &status, WNOHANG);
if(pid>0) {
child_procs_count--;
continue;
} else {
break;
}
}
FD_ZERO(&rfds);
FD_SET(ufd, &rfds);
if(child_procs_count < child_procs_max) {
FD_SET(lfd, &rfds);
}
maxfd = MAX(ufd, lfd) + 1;
timeout.tv_sec = 5;
timeout.tv_usec = 0;
result = select(maxfd, &rfds, 0, 0, &timeout);
if(result <= 0)
continue;
if(FD_ISSET(ufd, &rfds)) {
handle_updates(update_dgram);
}
if(FD_ISSET(lfd, &rfds)) {
link = link_accept(list_port, time(0) + 5);
if(link) {
if(fork_mode) {
pid_t pid = fork();
if(pid == 0) {
alarm(child_procs_timeout);
handle_query(link);
_exit(0);
} else if (pid>0) {
child_procs_count++;
}
} else {
handle_query(link);
}
link_close(link);
}
}
}
return 1;
}
int main(int argc, char *argv[]) {
char *host = CATALOG_HOST;
int port = CATALOG_PORT;
static const struct option long_options[] = {
{"catalog", required_argument, 0, 'c'},
{0,0,0,0}
};
signed int c;
while ((c = getopt_long(argc, argv, "c:", long_options, NULL)) > -1) {
switch (c) {
case 'c':
host = optarg;
break;
default:
show_help(argv[0]);
return EXIT_FAILURE;
}
}
struct datagram *d;
d = datagram_create(DATAGRAM_PORT_ANY);
if (!d) {
fatal("could not create datagram port!");
}
struct utsname name;
int cpus;
int uptime;
double load[3];
UINT64_T memory_total, memory_avail;
char owner[USERNAME_MAX];
uname(&name);
string_tolower(name.sysname);
string_tolower(name.machine);
string_tolower(name.release);
load_average_get(load);
cpus = load_average_get_cpus();
host_memory_info_get(&memory_avail, &memory_total);
uptime = uptime_get();
username_get(owner);
struct jx *j = jx_object(0);
jx_insert_string(j,"type","node");
jx_insert(j,jx_string("version"),jx_format("%d.%d.%d",CCTOOLS_VERSION_MAJOR, CCTOOLS_VERSION_MINOR, CCTOOLS_VERSION_MICRO));
jx_insert_string(j,"cpu",name.machine);
jx_insert_string(j,"opsys",name.sysname);
jx_insert_string(j,"opsysversion",name.release);
jx_insert_double(j,"load1",load[0]);
jx_insert_double(j,"load5",load[1]);
jx_insert_double(j,"load15",load[2]);
jx_insert_integer(j,"memory_total",memory_total);
jx_insert_integer(j,"memory_avail",memory_avail);
jx_insert_integer(j,"cpus",cpus);
jx_insert_integer(j,"uptime,",uptime);
jx_insert_string(j,"owner",owner);
int i;
for (i = optind; i < argc; i++) {
char *name;
char *value;
name = argv[i];
value = strchr(name, '=');
if (!value) {
fatal("invalid name/value pair = %s", name);
} else {
*value++ = 0;
}
jx_insert_string(j,name,value);
}
char *text = jx_print_string(j);
char address[DATAGRAM_ADDRESS_MAX];
if (domain_name_cache_lookup(host, address)) {
datagram_send(d, text, strlen(text), address, port);
} else {
fatal("unable to lookup address of host: %s", host);
}
jx_delete(j);
datagram_delete(d);
return EXIT_SUCCESS;
}
int main(int argc, char *argv[]) {
char *host = CATALOG_HOST;
int port = CATALOG_PORT;
static struct option long_options[] = {{"catalog", required_argument, 0, 'c'},
{0,0,0,0}};
signed int c;
while ((c = getopt_long(argc, argv, "c:", long_options, NULL)) > -1) {
switch (c) {
case 'c':
host = optarg;
break;
default:
show_help(argv[0]);
return EXIT_FAILURE;
}
}
struct datagram *d;
d = datagram_create(DATAGRAM_PORT_ANY);
if (!d) {
fatal("could not create datagram port!");
}
buffer_t B;
const char *text;
size_t text_size;
buffer_init(&B);
buffer_abortonfailure(&B, 1);
struct utsname name;
int cpus;
int uptime;
double load[3];
UINT64_T memory_total, memory_avail;
char owner[USERNAME_MAX];
uname(&name);
string_tolower(name.sysname);
string_tolower(name.machine);
string_tolower(name.release);
load_average_get(load);
cpus = load_average_get_cpus();
memory_info_get(&memory_avail, &memory_total);
uptime = uptime_get();
username_get(owner);
buffer_printf(&B, "type %s\nversion %d.%d.%s\ncpu %s\nopsys %s\nopsysversion %s\nload1 %0.02lf\nload5 %0.02lf\nload15 %0.02lf\nmemory_total %llu\nmemory_avail %llu\ncpus %d\nuptime %d\nowner %s\n",
DEFAULT_TYPE,
CCTOOLS_VERSION_MAJOR, CCTOOLS_VERSION_MINOR, CCTOOLS_VERSION_MICRO,
name.machine,
name.sysname,
name.release,
load[0],
load[1],
load[2],
(unsigned long long) memory_total,
(unsigned long long) memory_avail,
cpus,
uptime,
owner
);
int i;
for (i = optind; i < argc; i++) {
char *name;
char *value;
name = argv[i];
value = strchr(name, '=');
if (!value) {
fatal("invalid name/value pair = %s", name);
} else {
*value++ = 0;
}
buffer_printf(&B, "%s %s\n", name, value);
}
text = buffer_tostring(&B, &text_size);
char address[DATAGRAM_ADDRESS_MAX];
if (domain_name_cache_lookup(host, address)) {
datagram_send(d, text, text_size, address, port);
} else {
fatal("unable to lookup address of host: %s", host);
}
buffer_free(&B);
datagram_delete(d);
return EXIT_SUCCESS;
}
