#include "lib/tdb_wrap/tdb_wrap.h"
#include "lib/util/dlinklist.h"
#include "lib/util/debug.h"
-#include "lib/util/samba_util.h"
+#include "lib/util/time.h"
+#include "lib/util/blocking.h"
+#include "lib/util/become_daemon.h"
-#include "ctdb_version.h"
+#include "version.h"
#include "ctdb_private.h"
#include "ctdb_client.h"
-#include "ctdb_logging.h"
#include "common/rb_tree.h"
#include "common/reqid.h"
#include "common/system.h"
#include "common/common.h"
+#include "common/logging.h"
+#include "common/pidfile.h"
+#include "common/sock_io.h"
struct ctdb_client_pid_list {
struct ctdb_client_pid_list *next, *prev;
};
const char *ctdbd_pidfile = NULL;
+static struct pidfile_context *ctdbd_pidfile_ctx = NULL;
static void daemon_incoming_packet(void *, struct ctdb_req_header *);
+static pid_t __ctdbd_pid;
+
static void print_exit_message(void)
{
- if (debug_extra != NULL && debug_extra[0] != '\0') {
- DEBUG(DEBUG_NOTICE,("CTDB %s shutting down\n", debug_extra));
- } else {
+ if (getpid() == __ctdbd_pid) {
DEBUG(DEBUG_NOTICE,("CTDB daemon shutting down\n"));
/* Wait a second to allow pending log messages to be flushed */
}
}
+#ifdef HAVE_GETRUSAGE
+
+struct cpu_check_threshold_data {
+ unsigned short percent;
+ struct timeval timeofday;
+ struct timeval ru_time;
+};
+
+static void ctdb_cpu_check_threshold(struct tevent_context *ev,
+ struct tevent_timer *te,
+ struct timeval tv,
+ void *private_data)
+{
+ struct ctdb_context *ctdb = talloc_get_type_abort(
+ private_data, struct ctdb_context);
+ uint32_t interval = 60;
+
+ static unsigned short threshold = 0;
+ static struct cpu_check_threshold_data prev = {
+ .percent = 0,
+ .timeofday = { .tv_sec = 0 },
+ .ru_time = { .tv_sec = 0 },
+ };
+
+ struct rusage usage;
+ struct cpu_check_threshold_data curr = {
+ .percent = 0,
+ };
+ int64_t ru_time_diff, timeofday_diff;
+ bool first;
+ int ret;
+
+ /*
+ * Cache the threshold so that we don't waste time checking
+ * the environment variable every time
+ */
+ if (threshold == 0) {
+ const char *t;
+
+ threshold = 90;
+
+ t = getenv("CTDB_TEST_CPU_USAGE_THRESHOLD");
+ if (t != NULL) {
+ int th;
+
+ th = atoi(t);
+ if (th <= 0 || th > 100) {
+ DBG_WARNING("Failed to parse env var: %s\n", t);
+ } else {
+ threshold = th;
+ }
+ }
+ }
+
+ ret = getrusage(RUSAGE_SELF, &usage);
+ if (ret != 0) {
+ DBG_WARNING("rusage() failed: %d\n", ret);
+ goto next;
+ }
+
+ /* Sum the system and user CPU usage */
+ curr.ru_time = timeval_sum(&usage.ru_utime, &usage.ru_stime);
+ curr.timeofday = tv;
+
+ first = timeval_is_zero(&prev.timeofday);
+ if (first) {
+ /* No previous values recorded so no calculation to do */
+ goto done;
+ }
+
+ timeofday_diff = usec_time_diff(&curr.timeofday, &prev.timeofday);
+ if (timeofday_diff <= 0) {
+ /*
+ * Time went backwards or didn't progress so no (sane)
+ * calculation can be done
+ */
+ goto done;
+ }
+
+ ru_time_diff = usec_time_diff(&curr.ru_time, &prev.ru_time);
+
+ curr.percent = ru_time_diff * 100 / timeofday_diff;
+
+ if (curr.percent >= threshold) {
+ /* Log only if the utilisation changes */
+ if (curr.percent != prev.percent) {
+ D_WARNING("WARNING: CPU utilisation %hu%% >= "
+ "threshold (%hu%%)\n",
+ curr.percent,
+ threshold);
+ }
+ } else {
+ /* Log if the utilisation falls below the threshold */
+ if (prev.percent >= threshold) {
+ D_WARNING("WARNING: CPU utilisation %hu%% < "
+ "threshold (%hu%%)\n",
+ curr.percent,
+ threshold);
+ }
+ }
+
+done:
+ prev = curr;
+
+next:
+ tevent_add_timer(ctdb->ev, ctdb,
+ timeval_current_ofs(interval, 0),
+ ctdb_cpu_check_threshold,
+ ctdb);
+}
+
+static void ctdb_start_cpu_check_threshold(struct ctdb_context *ctdb)
+{
+ tevent_add_timer(ctdb->ev, ctdb,
+ timeval_current(),
+ ctdb_cpu_check_threshold,
+ ctdb);
+}
+#endif /* HAVE_GETRUSAGE */
static void ctdb_time_tick(struct tevent_context *ev, struct tevent_timer *te,
struct timeval t, void *private_data)
/* start listening to timer ticks */
ctdb_start_time_tickd(ctdb);
+
+#ifdef HAVE_GETRUSAGE
+ ctdb_start_cpu_check_threshold(ctdb);
+#endif /* HAVE_GETRUSAGE */
}
static void ignore_signal(int signum)
return srvid_deregister(ctdb->srv, srvid, client);
}
-int daemon_check_srvids(struct ctdb_context *ctdb, TDB_DATA indata,
- TDB_DATA *outdata)
+void daemon_tunnel_handler(uint64_t tunnel_id, TDB_DATA data,
+ void *private_data)
{
- uint64_t *ids;
- int i, num_ids;
- uint8_t *results;
+ struct ctdb_client *client =
+ talloc_get_type_abort(private_data, struct ctdb_client);
+ struct ctdb_req_tunnel_old *c, *pkt;
+ size_t len;
- if ((indata.dsize % sizeof(uint64_t)) != 0) {
- DEBUG(DEBUG_ERR, ("Bad indata in daemon_check_srvids, "
- "size=%d\n", (int)indata.dsize));
- return -1;
+ pkt = (struct ctdb_req_tunnel_old *)data.dptr;
+
+ len = offsetof(struct ctdb_req_tunnel_old, data) + pkt->datalen;
+ c = ctdbd_allocate_pkt(client->ctdb, client->ctdb, CTDB_REQ_TUNNEL,
+ len, struct ctdb_req_tunnel_old);
+ if (c == NULL) {
+ DEBUG(DEBUG_ERR, ("Memory error in daemon_tunnel_handler\n"));
+ return;
}
- ids = (uint64_t *)indata.dptr;
- num_ids = indata.dsize / 8;
+ talloc_set_name_const(c, "req_tunnel packet");
- results = talloc_zero_array(outdata, uint8_t, (num_ids+7)/8);
- if (results == NULL) {
- DEBUG(DEBUG_ERR, ("talloc failed in daemon_check_srvids\n"));
- return -1;
- }
- for (i=0; i<num_ids; i++) {
- if (srvid_exists(ctdb->srv, ids[i]) == 0) {
- results[i/8] |= (1 << (i%8));
- }
- }
- outdata->dptr = (uint8_t *)results;
- outdata->dsize = talloc_get_size(results);
- return 0;
+ c->tunnel_id = tunnel_id;
+ c->flags = pkt->flags;
+ c->datalen = pkt->datalen;
+ memcpy(c->data, pkt->data, pkt->datalen);
+
+ daemon_queue_send(client, &c->hdr);
+
+ talloc_free(c);
}
/*
dfc->w->ctdb = ctdb_db->ctdb;
dfc->w->client_id = client->client_id;
- DLIST_ADD_END(dfq->deferred_calls, dfc, NULL);
+ DLIST_ADD_END(dfq->deferred_calls, dfc);
return 0;
}
DEBUG(DEBUG_ERR,(__location__ " ctdb_ltdb_unlock() failed with error %d\n", ret));
}
CTDB_DECREMENT_STAT(ctdb, pending_calls);
+ talloc_free(data.dptr);
return;
}
}
- /* Dont do READONLY if we dont have a tracking database */
- if ((c->flags & CTDB_WANT_READONLY) && !ctdb_db->readonly) {
+ /* Dont do READONLY if we don't have a tracking database */
+ if ((c->flags & CTDB_WANT_READONLY) && !ctdb_db_readonly(ctdb_db)) {
c->flags &= ~CTDB_WANT_READONLY;
}
static void daemon_request_control_from_client(struct ctdb_client *client,
struct ctdb_req_control_old *c);
+static void daemon_request_tunnel_from_client(struct ctdb_client *client,
+ struct ctdb_req_tunnel_old *c);
/* data contains a packet from the client */
static void daemon_incoming_packet(void *p, struct ctdb_req_header *hdr)
daemon_request_control_from_client(client, (struct ctdb_req_control_old *)hdr);
break;
+ case CTDB_REQ_TUNNEL:
+ CTDB_INCREMENT_STAT(ctdb, client.req_tunnel);
+ daemon_request_tunnel_from_client(client, (struct ctdb_req_tunnel_old *)hdr);
+ break;
+
default:
DEBUG(DEBUG_CRIT,(__location__ " daemon: unrecognized operation %u\n",
hdr->operation));
return;
}
hdr = (struct ctdb_req_header *)data;
- if (cnt != hdr->length) {
- ctdb_set_error(client->ctdb, "Bad header length %u expected %u\n in daemon",
- (unsigned)hdr->length, (unsigned)cnt);
- return;
- }
if (hdr->ctdb_magic != CTDB_MAGIC) {
ctdb_set_error(client->ctdb, "Non CTDB packet rejected\n");
- return;
+ goto err_out;
}
if (hdr->ctdb_version != CTDB_PROTOCOL) {
ctdb_set_error(client->ctdb, "Bad CTDB version 0x%x rejected in daemon\n", hdr->ctdb_version);
- return;
+ goto err_out;
}
DEBUG(DEBUG_DEBUG,(__location__ " client request %u of type %u length %u from "
/* it is the responsibility of the incoming packet function to free 'data' */
daemon_incoming_packet(client, hdr);
+ return;
+
+err_out:
+ TALLOC_FREE(data);
}
return 0;
}
+static int get_new_client_id(struct reqid_context *idr,
+ struct ctdb_client *client,
+ uint32_t *out)
+{
+ uint32_t client_id;
+
+ client_id = reqid_new(idr, client);
+ /*
+ * Some places in the code (e.g. ctdb_control_db_attach(),
+ * ctdb_control_db_detach()) assign a special meaning to
+ * client_id 0. The assumption is that if client_id is 0 then
+ * the control has come from another daemon. Therefore, we
+ * should never return client_id == 0.
+ */
+ if (client_id == 0) {
+ /*
+ * Don't leak ID 0. This is safe because the ID keeps
+ * increasing. A test will be added to ensure that
+ * this doesn't change.
+ */
+ reqid_remove(idr, 0);
+
+ client_id = reqid_new(idr, client);
+ }
+
+ if (client_id == REQID_INVALID) {
+ return EINVAL;
+ }
+
+ if (client_id == 0) {
+ /* Every other ID must have been used and we can't use 0 */
+ reqid_remove(idr, 0);
+ return EINVAL;
+ }
+
+ *out = client_id;
+ return 0;
+}
static void ctdb_accept_client(struct tevent_context *ev,
struct tevent_fd *fde, uint16_t flags,
struct ctdb_client *client;
struct ctdb_client_pid_list *client_pid;
pid_t peer_pid = 0;
+ int ret;
memset(&addr, 0, sizeof(addr));
len = sizeof(addr);
if (fd == -1) {
return;
}
+ smb_set_close_on_exec(fd);
+
+ ret = set_blocking(fd, false);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,
+ (__location__
+ " failed to set socket non-blocking (%s)\n",
+ strerror(errno)));
+ close(fd);
+ return;
+ }
- set_nonblocking(fd);
set_close_on_exec(fd);
DEBUG(DEBUG_DEBUG,(__location__ " Created SOCKET FD:%d to connected child\n", fd));
client->ctdb = ctdb;
client->fd = fd;
- client->client_id = reqid_new(ctdb->idr, client);
+
+ ret = get_new_client_id(ctdb->idr, client, &client->client_id);
+ if (ret != 0) {
+ DBG_ERR("Unable to get client ID (%d)\n", ret);
+ close(fd);
+ talloc_free(client);
+ return;
+ }
+
client->pid = peer_pid;
client_pid = talloc(client, struct ctdb_client_pid_list);
static int ux_socket_bind(struct ctdb_context *ctdb)
{
struct sockaddr_un addr;
+ int ret;
ctdb->daemon.sd = socket(AF_UNIX, SOCK_STREAM, 0);
if (ctdb->daemon.sd == -1) {
addr.sun_family = AF_UNIX;
strncpy(addr.sun_path, ctdb->daemon.name, sizeof(addr.sun_path)-1);
- /* First check if an old ctdbd might be running */
- if (connect(ctdb->daemon.sd,
- (struct sockaddr *)&addr, sizeof(addr)) == 0) {
- DEBUG(DEBUG_CRIT,
- ("Something is already listening on ctdb socket '%s'\n",
- ctdb->daemon.name));
- goto failed;
+ if (! sock_clean(ctdb->daemon.name)) {
+ return -1;
}
- /* Remove any old socket */
- unlink(ctdb->daemon.name);
-
set_close_on_exec(ctdb->daemon.sd);
- set_nonblocking(ctdb->daemon.sd);
+
+ ret = set_blocking(ctdb->daemon.sd, false);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,
+ (__location__
+ " failed to set socket non-blocking (%s)\n",
+ strerror(errno)));
+ goto failed;
+ }
if (bind(ctdb->daemon.sd, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
DEBUG(DEBUG_CRIT,("Unable to bind on ctdb socket '%s'\n", ctdb->daemon.name));
goto failed;
}
+ DEBUG(DEBUG_NOTICE, ("Listening to ctdb socket %s\n",
+ ctdb->daemon.name));
return 0;
failed:
/* do we start out in DISABLED mode? */
if (ctdb->start_as_disabled != 0) {
- DEBUG(DEBUG_NOTICE, ("This node is configured to start in DISABLED state\n"));
+ DEBUG(DEBUG_ERR,
+ ("This node is configured to start in DISABLED state\n"));
ctdb->nodes[ctdb->pnn]->flags |= NODE_FLAGS_DISABLED;
}
/* do we start out in STOPPED mode? */
if (ctdb->start_as_stopped != 0) {
- DEBUG(DEBUG_NOTICE, ("This node is configured to start in STOPPED state\n"));
+ DEBUG(DEBUG_ERR,
+ ("This node is configured to start in STOPPED state\n"));
ctdb->nodes[ctdb->pnn]->flags |= NODE_FLAGS_STOPPED;
}
}
}
ctdb_run_notification_script(ctdb, "setup");
- /* tell all other nodes we've just started up */
- ctdb_daemon_send_control(ctdb, CTDB_BROADCAST_ALL,
- 0, CTDB_CONTROL_STARTUP, 0,
- CTDB_CTRL_FLAG_NOREPLY,
- tdb_null, NULL, NULL);
-
/* Start the recovery daemon */
if (ctdb_start_recoverd(ctdb) != 0) {
DEBUG(DEBUG_ALERT,("Failed to start recovery daemon\n"));
static struct timeval tevent_before_wait_ts;
static struct timeval tevent_after_wait_ts;
+static void ctdb_tevent_trace_init(void)
+{
+ struct timeval now;
+
+ now = timeval_current();
+
+ tevent_before_wait_ts = now;
+ tevent_after_wait_ts = now;
+}
+
static void ctdb_tevent_trace(enum tevent_trace_point tp,
void *private_data)
{
switch (tp) {
case TEVENT_TRACE_BEFORE_WAIT:
- if (!timeval_is_zero(&tevent_after_wait_ts)) {
- diff = timeval_until(&tevent_after_wait_ts, &now);
- if (diff.tv_sec > 3) {
- DEBUG(DEBUG_ERR,
- ("Handling event took %ld seconds!\n",
- (long)diff.tv_sec));
- }
+ diff = timeval_until(&tevent_after_wait_ts, &now);
+ if (diff.tv_sec > 3) {
+ DEBUG(DEBUG_ERR,
+ ("Handling event took %ld seconds!\n",
+ (long)diff.tv_sec));
}
tevent_before_wait_ts = now;
break;
case TEVENT_TRACE_AFTER_WAIT:
- if (!timeval_is_zero(&tevent_before_wait_ts)) {
- diff = timeval_until(&tevent_before_wait_ts, &now);
- if (diff.tv_sec > 3) {
- DEBUG(DEBUG_CRIT,
- ("No event for %ld seconds!\n",
- (long)diff.tv_sec));
- }
+ diff = timeval_until(&tevent_before_wait_ts, &now);
+ if (diff.tv_sec > 3) {
+ DEBUG(DEBUG_ERR,
+ ("No event for %ld seconds!\n",
+ (long)diff.tv_sec));
}
tevent_after_wait_ts = now;
break;
static void ctdb_remove_pidfile(void)
{
- /* Only the main ctdbd's PID matches the SID */
- if (ctdbd_pidfile != NULL && getsid(0) == getpid()) {
- if (unlink(ctdbd_pidfile) == 0) {
- DEBUG(DEBUG_NOTICE, ("Removed PID file %s\n",
- ctdbd_pidfile));
- } else {
- DEBUG(DEBUG_WARNING, ("Failed to Remove PID file %s\n",
- ctdbd_pidfile));
- }
- }
+ TALLOC_FREE(ctdbd_pidfile_ctx);
}
-static void ctdb_create_pidfile(pid_t pid)
+static void ctdb_create_pidfile(TALLOC_CTX *mem_ctx)
{
if (ctdbd_pidfile != NULL) {
- FILE *fp;
-
- fp = fopen(ctdbd_pidfile, "w");
- if (fp == NULL) {
- DEBUG(DEBUG_ALERT,
- ("Failed to open PID file %s\n", ctdbd_pidfile));
+ int ret = pidfile_context_create(mem_ctx, ctdbd_pidfile,
+ &ctdbd_pidfile_ctx);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,
+ ("Failed to create PID file %s\n",
+ ctdbd_pidfile));
exit(11);
}
- fprintf(fp, "%d\n", pid);
- fclose(fp);
DEBUG(DEBUG_NOTICE, ("Created PID file %s\n", ctdbd_pidfile));
atexit(ctdb_remove_pidfile);
}
static void ctdb_initialise_vnn_map(struct ctdb_context *ctdb)
{
- int i, j, count;
+ unsigned int i, j, count;
/* initialize the vnn mapping table, skipping any deleted nodes */
ctdb->vnn_map = talloc(ctdb, struct ctdb_vnn_map);
int res, ret = -1;
struct tevent_fd *fde;
- /* create a unix domain stream socket to listen to */
- res = ux_socket_bind(ctdb);
- if (res!=0) {
- DEBUG(DEBUG_ALERT,("Cannot continue. Exiting!\n"));
- exit(10);
- }
-
- if (do_fork && fork()) {
- return 0;
- }
-
- tdb_reopen_all(false);
+ become_daemon(do_fork, !do_fork, false);
- if (do_fork) {
- if (setsid() == -1) {
- ctdb_die(ctdb, "Failed to setsid()\n");
- }
- close(0);
- if (open("/dev/null", O_RDONLY) != 0) {
- DEBUG(DEBUG_ALERT,(__location__ " Failed to setup stdin on /dev/null\n"));
- exit(11);
- }
- }
ignore_signal(SIGPIPE);
ignore_signal(SIGUSR1);
ctdb->ctdbd_pid = getpid();
DEBUG(DEBUG_ERR, ("Starting CTDBD (Version %s) as PID: %u\n",
- CTDB_VERSION_STRING, ctdb->ctdbd_pid));
- ctdb_create_pidfile(ctdb->ctdbd_pid);
+ SAMBA_VERSION_STRING, ctdb->ctdbd_pid));
+ ctdb_create_pidfile(ctdb);
+
+ /* create a unix domain stream socket to listen to */
+ res = ux_socket_bind(ctdb);
+ if (res!=0) {
+ DEBUG(DEBUG_ALERT,("Cannot continue. Exiting!\n"));
+ exit(10);
+ }
/* Make sure we log something when the daemon terminates.
* This must be the first exit handler to run (so the last to
* be registered.
*/
+ __ctdbd_pid = getpid();
atexit(print_exit_message);
if (ctdb->do_setsched) {
}
ctdb->ev = tevent_context_init(NULL);
- tevent_loop_allow_nesting(ctdb->ev);
- tevent_set_trace_callback(ctdb->ev, ctdb_tevent_trace, ctdb);
- ret = ctdb_init_tevent_logging(ctdb);
- if (ret != 0) {
- DEBUG(DEBUG_ALERT,("Failed to initialize TEVENT logging\n"));
+ if (ctdb->ev == NULL) {
+ DEBUG(DEBUG_ALERT,("tevent_context_init() failed\n"));
exit(1);
}
+ tevent_loop_allow_nesting(ctdb->ev);
+ ctdb_tevent_trace_init();
+ tevent_set_trace_callback(ctdb->ev, ctdb_tevent_trace, ctdb);
/* set up a handler to pick up sigchld */
if (ctdb_init_sigchld(ctdb) == NULL) {
exit(1);
}
- ctdb_set_child_logging(ctdb);
+ if (do_fork) {
+ ctdb_set_child_logging(ctdb);
+ }
+ TALLOC_FREE(ctdb->srv);
if (srvid_init(ctdb, &ctdb->srv) != 0) {
DEBUG(DEBUG_CRIT,("Failed to setup message srvid context\n"));
exit(1);
}
+ TALLOC_FREE(ctdb->tunnels);
+ if (srvid_init(ctdb, &ctdb->tunnels) != 0) {
+ DEBUG(DEBUG_ERR, ("Failed to setup tunnels context\n"));
+ exit(1);
+ }
+
/* initialize statistics collection */
ctdb_statistics_init(ctdb);
/* force initial recovery for election */
ctdb->recovery_mode = CTDB_RECOVERY_ACTIVE;
+ if (ctdb_start_eventd(ctdb) != 0) {
+ DEBUG(DEBUG_ERR, ("Failed to start event daemon\n"));
+ exit(1);
+ }
+
ctdb_set_runstate(ctdb, CTDB_RUNSTATE_INIT);
ret = ctdb_event_script(ctdb, CTDB_EVENT_INIT);
if (ret != 0) {
initialise_node_flags(ctdb);
- if (ctdb->public_addresses_file) {
- ret = ctdb_set_public_addresses(ctdb, true);
- if (ret == -1) {
- DEBUG(DEBUG_ALERT,("Unable to setup public address list\n"));
- exit(1);
- }
+ ret = ctdb_set_public_addresses(ctdb, true);
+ if (ret == -1) {
+ D_ERR("Unable to setup public IP addresses\n");
+ exit(1);
}
ctdb_initialise_vnn_map(ctdb);
talloc_free(tmp_ctx);
}
+static void daemon_request_tunnel_from_client(struct ctdb_client *client,
+ struct ctdb_req_tunnel_old *c)
+{
+ TDB_DATA data;
+ int ret;
+
+ if (! ctdb_validate_pnn(client->ctdb, c->hdr.destnode)) {
+ DEBUG(DEBUG_ERR, ("Invalid destination 0x%x\n",
+ c->hdr.destnode));
+ return;
+ }
+
+ ret = srvid_exists(client->ctdb->tunnels, c->tunnel_id, NULL);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR,
+ ("tunnel id 0x%"PRIx64" not registered, dropping pkt\n",
+ c->tunnel_id));
+ return;
+ }
+
+ data = (TDB_DATA) {
+ .dsize = c->datalen,
+ .dptr = &c->data[0],
+ };
+
+ ret = ctdb_daemon_send_tunnel(client->ctdb, c->hdr.destnode,
+ c->tunnel_id, c->flags, data);
+ if (ret != 0) {
+ DEBUG(DEBUG_ERR, ("Failed to set tunnel to remote note %u\n",
+ c->hdr.destnode));
+ }
+}
+
/*
register a call function
*/
nl->ctdb = ctdb;
nl->srvid = notify->srvid;
nl->data.dsize = notify->len;
- nl->data.dptr = talloc_size(nl, nl->data.dsize);
+ nl->data.dptr = talloc_memdup(nl, notify->notify_data,
+ nl->data.dsize);
CTDB_NO_MEMORY(ctdb, nl->data.dptr);
- memcpy(nl->data.dptr, notify->notify_data, nl->data.dsize);
DLIST_ADD(client->notify, nl);
talloc_set_destructor(nl, ctdb_client_notify_destructor);
int32_t ctdb_control_deregister_notify(struct ctdb_context *ctdb, uint32_t client_id, TDB_DATA indata)
{
- struct ctdb_client_notify_deregister *notify = (struct ctdb_client_notify_deregister *)indata.dptr;
+ uint64_t srvid = *(uint64_t *)indata.dptr;
struct ctdb_client *client = reqid_find(ctdb->idr, client_id, struct ctdb_client);
struct ctdb_client_notify_list *nl;
- DEBUG(DEBUG_INFO,("Deregister srvid %llu for client %d\n", (unsigned long long)notify->srvid, client_id));
+ DEBUG(DEBUG_INFO,("Deregister srvid %llu for client %d\n", (unsigned long long)srvid, client_id));
if (client == NULL) {
DEBUG(DEBUG_ERR,(__location__ " Could not find client parent structure. You can not send this control to a remote node\n"));
}
for(nl=client->notify; nl; nl=nl->next) {
- if (nl->srvid == notify->srvid) {
+ if (nl->srvid == srvid) {
break;
}
}
if (nl == NULL) {
- DEBUG(DEBUG_ERR,(__location__ " No notification for srvid:%llu found for this client\n", (unsigned long long)notify->srvid));
+ DEBUG(DEBUG_ERR,(__location__ " No notification for srvid:%llu found for this client\n", (unsigned long long)srvid));
return -1;
}
/* This control is used by samba when probing if a process (of a samba daemon)
exists on the node.
Samba does this when it needs/wants to check if a subrecord in one of the
- databases is still valied, or if it is stale and can be removed.
+ databases is still valid, or if it is stale and can be removed.
If the node is in unhealthy or stopped state we just kill of the samba
- process holding htis sub-record and return to the calling samba that
+ process holding this sub-record and return to the calling samba that
the process does not exist.
This allows us to forcefully recall subrecords registered by samba processes
on banned and stopped nodes.
{
struct ctdb_client *client;
- if (ctdb->nodes[ctdb->pnn]->flags & (NODE_FLAGS_BANNED|NODE_FLAGS_STOPPED)) {
- client = ctdb_find_client_by_pid(ctdb, pid);
- if (client != NULL) {
- DEBUG(DEBUG_NOTICE,(__location__ " Killing client with pid:%d on banned/stopped node\n", (int)pid));
- talloc_free(client);
- }
+ client = ctdb_find_client_by_pid(ctdb, pid);
+ if (client == NULL) {
+ return -1;
+ }
+
+ if (ctdb->nodes[ctdb->pnn]->flags & NODE_FLAGS_INACTIVE) {
+ DEBUG(DEBUG_NOTICE,
+ ("Killing client with pid:%d on banned/stopped node\n",
+ (int)pid));
+ talloc_free(client);
return -1;
}
return kill(pid, 0);
}
+int32_t ctdb_control_check_pid_srvid(struct ctdb_context *ctdb,
+ TDB_DATA indata)
+{
+ struct ctdb_client_pid_list *client_pid;
+ pid_t pid;
+ uint64_t srvid;
+ int ret;
+
+ pid = *(pid_t *)indata.dptr;
+ srvid = *(uint64_t *)(indata.dptr + sizeof(pid_t));
+
+ for (client_pid = ctdb->client_pids;
+ client_pid != NULL;
+ client_pid = client_pid->next) {
+ if (client_pid->pid == pid) {
+ ret = srvid_exists(ctdb->srv, srvid,
+ client_pid->client);
+ if (ret == 0) {
+ return 0;
+ }
+ }
+ }
+
+ return -1;
+}
+
int ctdb_control_getnodesfile(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata, TDB_DATA *outdata)
{
struct ctdb_node_map_old *node_map = NULL;
return;
}
- DEBUG(DEBUG_NOTICE,("Shutdown sequence commencing.\n"));
+ DEBUG(DEBUG_ERR,("Shutdown sequence commencing.\n"));
ctdb_set_runstate(ctdb, CTDB_RUNSTATE_SHUTDOWN);
ctdb_stop_recoverd(ctdb);
ctdb_stop_keepalive(ctdb);
ctdb_stop_monitoring(ctdb);
ctdb_release_all_ips(ctdb);
ctdb_event_script(ctdb, CTDB_EVENT_SHUTDOWN);
- if (ctdb->methods != NULL) {
+ ctdb_stop_eventd(ctdb);
+ if (ctdb->methods != NULL && ctdb->methods->shutdown != NULL) {
ctdb->methods->shutdown(ctdb);
}
- DEBUG(DEBUG_NOTICE,("Shutdown sequence complete, exiting.\n"));
+ DEBUG(DEBUG_ERR,("Shutdown sequence complete, exiting.\n"));
exit(exit_code);
}
+
+/* When forking the main daemon and the child process needs to connect
+ * back to the daemon as a client process, this function can be used
+ * to change the ctdb context from daemon into client mode. The child
+ * process must be created using ctdb_fork() and not fork() -
+ * ctdb_fork() does some necessary housekeeping.
+ */
+int switch_from_server_to_client(struct ctdb_context *ctdb)
+{
+ int ret;
+
+ /* get a new event context */
+ ctdb->ev = tevent_context_init(ctdb);
+ if (ctdb->ev == NULL) {
+ DEBUG(DEBUG_ALERT,("tevent_context_init() failed\n"));
+ exit(1);
+ }
+ tevent_loop_allow_nesting(ctdb->ev);
+
+ /* Connect to main CTDB daemon */
+ ret = ctdb_socket_connect(ctdb);
+ if (ret != 0) {
+ DEBUG(DEBUG_ALERT, (__location__ " Failed to init ctdb client\n"));
+ return -1;
+ }
+
+ ctdb->can_send_controls = true;
+
+ return 0;
+}