along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#include "includes.h"
-#include "lib/events/events.h"
+#include "lib/tevent/tevent.h"
#include "lib/tdb/include/tdb.h"
#include "system/time.h"
#include "system/network.h"
/*
lock all databases - mark only
*/
-static int ctdb_lock_all_databases_mark(struct ctdb_context *ctdb)
+static int ctdb_lock_all_databases_mark(struct ctdb_context *ctdb, uint32_t priority)
{
struct ctdb_db_context *ctdb_db;
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
+
+ if ((priority < 1) || (priority > NUM_DB_PRIORITIES)) {
+ DEBUG(DEBUG_ERR,(__location__ " Illegal priority when trying to mark all databases Prio:%u\n", priority));
+ return -1;
+ }
+
+ if (ctdb->freeze_mode[priority] != CTDB_FREEZE_FROZEN) {
DEBUG(DEBUG_ERR,("Attempt to mark all databases locked when not frozen\n"));
return -1;
}
+ /* The dual loop is a woraround for older versions of samba
+ that does not yet support the set-db-priority/lock order
+ call. So that we get basic deadlock avoiidance also for
+ these old versions of samba.
+ This code will be removed in the future.
+ */
for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ if (ctdb_db->priority != priority) {
+ continue;
+ }
+ if (strstr(ctdb_db->db_name, "notify") != NULL) {
+ continue;
+ }
+ if (tdb_transaction_write_lock_mark(ctdb_db->ltdb->tdb) != 0) {
+ return -1;
+ }
if (tdb_lockall_mark(ctdb_db->ltdb->tdb) != 0) {
+ tdb_transaction_write_lock_unmark(ctdb_db->ltdb->tdb);
+ return -1;
+ }
+ }
+ for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ if (ctdb_db->priority != priority) {
+ continue;
+ }
+ if (strstr(ctdb_db->db_name, "notify") == NULL) {
+ continue;
+ }
+ if (tdb_transaction_write_lock_mark(ctdb_db->ltdb->tdb) != 0) {
+ return -1;
+ }
+ if (tdb_lockall_mark(ctdb_db->ltdb->tdb) != 0) {
+ tdb_transaction_write_lock_unmark(ctdb_db->ltdb->tdb);
return -1;
}
}
/*
lock all databases - unmark only
*/
-static int ctdb_lock_all_databases_unmark(struct ctdb_context *ctdb)
+static int ctdb_lock_all_databases_unmark(struct ctdb_context *ctdb, uint32_t priority)
{
struct ctdb_db_context *ctdb_db;
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
+
+ if ((priority < 1) || (priority > NUM_DB_PRIORITIES)) {
+ DEBUG(DEBUG_ERR,(__location__ " Illegal priority when trying to mark all databases Prio:%u\n", priority));
+ return -1;
+ }
+
+ if (ctdb->freeze_mode[priority] != CTDB_FREEZE_FROZEN) {
DEBUG(DEBUG_ERR,("Attempt to unmark all databases locked when not frozen\n"));
return -1;
}
for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ if (ctdb_db->priority != priority) {
+ continue;
+ }
+ tdb_transaction_write_lock_unmark(ctdb_db->ltdb->tdb);
if (tdb_lockall_unmark(ctdb_db->ltdb->tdb) != 0) {
return -1;
}
ctdb_control_setvnnmap(struct ctdb_context *ctdb, uint32_t opcode, TDB_DATA indata, TDB_DATA *outdata)
{
struct ctdb_vnn_map_wire *map = (struct ctdb_vnn_map_wire *)indata.dptr;
+ int i;
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
- DEBUG(DEBUG_ERR,("Attempt to set vnnmap when not frozen\n"));
- return -1;
+ for(i=1; i<=NUM_DB_PRIORITIES; i++) {
+ if (ctdb->freeze_mode[i] != CTDB_FREEZE_FROZEN) {
+ DEBUG(DEBUG_ERR,("Attempt to set vnnmap when not frozen\n"));
+ return -1;
+ }
}
talloc_free(ctdb->vnn_map);
}
params->pulldata = talloc_realloc_size(NULL, params->pulldata, rec->length + params->len);
if (params->pulldata == NULL) {
- DEBUG(DEBUG_ERR,(__location__ " Failed to expand pulldb_data to %u (%u records)\n",
- rec->length + params->len, params->pulldata->count));
- params->failed = true;
- return -1;
+ DEBUG(DEBUG_CRIT,(__location__ " Failed to expand pulldb_data to %u\n", rec->length + params->len));
+ ctdb_fatal(params->ctdb, "failed to allocate memory for recovery. shutting down\n");
}
params->pulldata->count++;
memcpy(params->len+(uint8_t *)params->pulldata, rec, rec->length);
struct pulldb_data params;
struct ctdb_marshall_buffer *reply;
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
- DEBUG(DEBUG_DEBUG,("rejecting ctdb_control_pull_db when not frozen\n"));
- return -1;
- }
-
pull = (struct ctdb_control_pulldb *)indata.dptr;
ctdb_db = find_ctdb_db(ctdb, pull->db_id);
return -1;
}
+ if (ctdb->freeze_mode[ctdb_db->priority] != CTDB_FREEZE_FROZEN) {
+ DEBUG(DEBUG_DEBUG,("rejecting ctdb_control_pull_db when not frozen\n"));
+ return -1;
+ }
+
reply = talloc_zero(outdata, struct ctdb_marshall_buffer);
CTDB_NO_MEMORY(ctdb, reply);
params.len = offsetof(struct ctdb_marshall_buffer, data);
params.failed = false;
- if (ctdb_lock_all_databases_mark(ctdb) != 0) {
+ if (ctdb_db->unhealthy_reason) {
+ /* this is just a warning, as the tdb should be empty anyway */
+ DEBUG(DEBUG_WARNING,("db(%s) unhealty in ctdb_control_pull_db: %s\n",
+ ctdb_db->db_name, ctdb_db->unhealthy_reason));
+ }
+
+ if (ctdb_lock_all_databases_mark(ctdb, ctdb_db->priority) != 0) {
DEBUG(DEBUG_ERR,(__location__ " Failed to get lock on entired db - failing\n"));
return -1;
}
if (tdb_traverse_read(ctdb_db->ltdb->tdb, traverse_pulldb, ¶ms) == -1) {
DEBUG(DEBUG_ERR,(__location__ " Failed to get traverse db '%s'\n", ctdb_db->db_name));
- ctdb_lock_all_databases_unmark(ctdb);
+ ctdb_lock_all_databases_unmark(ctdb, ctdb_db->priority);
talloc_free(params.pulldata);
return -1;
}
- ctdb_lock_all_databases_unmark(ctdb);
+ ctdb_lock_all_databases_unmark(ctdb, ctdb_db->priority);
outdata->dptr = (uint8_t *)params.pulldata;
outdata->dsize = params.len;
int i, ret;
struct ctdb_rec_data *rec;
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
- DEBUG(DEBUG_DEBUG,("rejecting ctdb_control_push_db when not frozen\n"));
- return -1;
- }
-
if (indata.dsize < offsetof(struct ctdb_marshall_buffer, data)) {
DEBUG(DEBUG_ERR,(__location__ " invalid data in pulldb reply\n"));
return -1;
return -1;
}
- if (ctdb_lock_all_databases_mark(ctdb) != 0) {
+ if (ctdb->freeze_mode[ctdb_db->priority] != CTDB_FREEZE_FROZEN) {
+ DEBUG(DEBUG_DEBUG,("rejecting ctdb_control_push_db when not frozen\n"));
+ return -1;
+ }
+
+ if (ctdb_lock_all_databases_mark(ctdb, ctdb_db->priority) != 0) {
DEBUG(DEBUG_ERR,(__location__ " Failed to get lock on entired db - failing\n"));
return -1;
}
DEBUG(DEBUG_DEBUG,("finished push of %u records for dbid 0x%x\n",
reply->count, reply->db_id));
- ctdb_lock_all_databases_unmark(ctdb);
+ ctdb_lock_all_databases_unmark(ctdb, ctdb_db->priority);
return 0;
failed:
- ctdb_lock_all_databases_unmark(ctdb);
+ ctdb_lock_all_databases_unmark(ctdb, ctdb_db->priority);
return -1;
}
struct ctdb_control_set_dmaster *p = (struct ctdb_control_set_dmaster *)indata.dptr;
struct ctdb_db_context *ctdb_db;
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
- DEBUG(DEBUG_DEBUG,("rejecting ctdb_control_set_dmaster when not frozen\n"));
- return -1;
- }
-
ctdb_db = find_ctdb_db(ctdb, p->db_id);
if (!ctdb_db) {
DEBUG(DEBUG_ERR,(__location__ " Unknown db 0x%08x\n", p->db_id));
return -1;
}
- if (ctdb_lock_all_databases_mark(ctdb) != 0) {
+ if (ctdb->freeze_mode[ctdb_db->priority] != CTDB_FREEZE_FROZEN) {
+ DEBUG(DEBUG_DEBUG,("rejecting ctdb_control_set_dmaster when not frozen\n"));
+ return -1;
+ }
+
+ if (ctdb_lock_all_databases_mark(ctdb, ctdb_db->priority) != 0) {
DEBUG(DEBUG_ERR,(__location__ " Failed to get lock on entired db - failing\n"));
return -1;
}
tdb_traverse(ctdb_db->ltdb->tdb, traverse_setdmaster, &p->dmaster);
- ctdb_lock_all_databases_unmark(ctdb);
+ ctdb_lock_all_databases_unmark(ctdb, ctdb_db->priority);
return 0;
}
{
double l = timeval_elapsed(&state->start_time);
- ctdb_reclock_latency(state->ctdb, "daemon reclock", &state->ctdb->statistics.reclock.ctdbd, l);
-
+ CTDB_UPDATE_RECLOCK_LATENCY(state->ctdb, "daemon reclock", reclock.ctdbd, l);
+
+ if (state->fd[0] != -1) {
+ state->fd[0] = -1;
+ }
+ if (state->fd[1] != -1) {
+ state->fd[1] = -1;
+ }
kill(state->child, SIGKILL);
return 0;
}
state->ctdb->recovery_mode = state->recmode;
+ /* release any deferred attach calls from clients */
+ if (state->recmode == CTDB_RECOVERY_NORMAL) {
+ ctdb_process_deferred_attach(state->ctdb);
+ }
+
ctdb_request_control_reply(state->ctdb, state->c, NULL, 0, NULL);
talloc_free(state);
return;
ctdb_release_all_ips(ctdb);
}
+/*
+ * Set up an event to drop all public ips if we remain in recovery for too
+ * long
+ */
+int ctdb_deferred_drop_all_ips(struct ctdb_context *ctdb)
+{
+ if (ctdb->release_ips_ctx != NULL) {
+ talloc_free(ctdb->release_ips_ctx);
+ }
+ ctdb->release_ips_ctx = talloc_new(ctdb);
+ CTDB_NO_MEMORY(ctdb, ctdb->release_ips_ctx);
+
+ event_add_timed(ctdb->ev, ctdb->release_ips_ctx, timeval_current_ofs(ctdb->tunable.recovery_drop_all_ips, 0), ctdb_drop_all_ips_event, ctdb);
+ return 0;
+}
+
/*
set the recovery mode
*/
const char **errormsg)
{
uint32_t recmode = *(uint32_t *)indata.dptr;
- int ret;
+ int i, ret;
struct ctdb_set_recmode_state *state;
pid_t parent = getpid();
talloc_free(ctdb->release_ips_ctx);
ctdb->release_ips_ctx = NULL;
} else {
- talloc_free(ctdb->release_ips_ctx);
- ctdb->release_ips_ctx = talloc_new(ctdb);
- CTDB_NO_MEMORY(ctdb, ctdb->release_ips_ctx);
-
- event_add_timed(ctdb->ev, ctdb->release_ips_ctx, timeval_current_ofs(ctdb->tunable.recovery_drop_all_ips, 0), ctdb_drop_all_ips_event, ctdb);
- }
-
-
- if (ctdb->freeze_mode != CTDB_FREEZE_FROZEN) {
- DEBUG(DEBUG_ERR,("Attempt to change recovery mode to %u when not frozen\n",
- recmode));
- (*errormsg) = "Cannot change recovery mode while not frozen";
- return -1;
+ if (ctdb_deferred_drop_all_ips(ctdb) != 0) {
+ DEBUG(DEBUG_ERR,("Failed to set up deferred drop all ips\n"));
+ }
}
if (recmode != ctdb->recovery_mode) {
/* some special handling when ending recovery mode */
- /* force the databased to thaw */
- if (ctdb->freeze_handle) {
- ctdb_control_thaw(ctdb);
+ /* force the databases to thaw */
+ for (i=1; i<=NUM_DB_PRIORITIES; i++) {
+ if (ctdb->freeze_handles[i] != NULL) {
+ ctdb_control_thaw(ctdb, i);
+ }
}
state = talloc(ctdb, struct ctdb_set_recmode_state);
CTDB_NO_MEMORY(ctdb, state);
state->start_time = timeval_current();
+ state->fd[0] = -1;
+ state->fd[1] = -1;
+
+ /* release any deferred attach calls from clients */
+ if (recmode == CTDB_RECOVERY_NORMAL) {
+ ctdb_process_deferred_attach(ctdb);
+ }
if (ctdb->tunable.verify_recovery_lock == 0) {
/* dont need to verify the reclock file */
char cc = 0;
close(state->fd[0]);
+ debug_extra = talloc_asprintf(NULL, "set_recmode:");
/* we should not be able to get the lock on the reclock file,
as it should be held by the recovery master
*/
_exit(0);
}
close(state->fd[1]);
+ set_close_on_exec(state->fd[0]);
+
+ state->fd[1] = -1;
talloc_set_destructor(state, set_recmode_destructor);
+ DEBUG(DEBUG_DEBUG, (__location__ " Created PIPE FD:%d for setrecmode\n", state->fd[0]));
+
state->te = event_add_timed(ctdb->ev, state, timeval_current_ofs(5, 0),
ctdb_set_recmode_timeout, state);
state->fde = event_add_fd(ctdb->ev, state, state->fd[0],
- EVENT_FD_READ|EVENT_FD_AUTOCLOSE,
+ EVENT_FD_READ,
set_recmode_handler,
(void *)state);
talloc_free(state);
return -1;
}
+ tevent_fd_set_auto_close(state->fde);
state->ctdb = ctdb;
state->recmode = recmode;
}
if (ctdb->recovery_lock_fd != -1) {
close(ctdb->recovery_lock_fd);
+ ctdb->recovery_lock_fd = -1;
}
+
ctdb->recovery_lock_fd = open(ctdb->recovery_lock_file, O_RDWR|O_CREAT, 0600);
if (ctdb->recovery_lock_fd == -1) {
DEBUG(DEBUG_ERR,("ctdb_recovery_lock: Unable to open %s - (%s)\n",
}
if (keep) {
- DEBUG(DEBUG_ERR, ("Recovery lock taken successfully\n"));
+ DEBUG(DEBUG_NOTICE, ("Recovery lock taken successfully\n"));
}
DEBUG(DEBUG_NOTICE,("ctdb_recovery_lock: Got recovery lock on '%s'\n", ctdb->recovery_lock_file));
struct recovery_callback_state *state = talloc_get_type(p, struct recovery_callback_state);
ctdb_enable_monitoring(ctdb);
+ CTDB_INCREMENT_STAT(ctdb, num_recoveries);
if (status != 0) {
DEBUG(DEBUG_ERR,(__location__ " recovered event script failed (status %d)\n", status));
+ if (status == -ETIME) {
+ ctdb_ban_self(ctdb);
+ }
}
ctdb_request_control_reply(ctdb, state->c, NULL, status, NULL);
DEBUG(DEBUG_NOTICE,("Recovery has finished\n"));
+ ctdb_persistent_finish_trans3_commits(ctdb);
+
state = talloc(ctdb, struct recovery_callback_state);
CTDB_NO_MEMORY(ctdb, state);
- state->c = talloc_steal(state, c);
+ state->c = c;
ctdb_disable_monitoring(ctdb);
- ret = ctdb_event_script_callback(ctdb,
- timeval_current_ofs(ctdb->tunable.script_timeout, 0),
- state,
+ ret = ctdb_event_script_callback(ctdb, state,
ctdb_end_recovery_callback,
- state, "recovered");
+ state,
+ false,
+ CTDB_EVENT_RECOVERED, "%s", "");
if (ret != 0) {
ctdb_enable_monitoring(ctdb);
}
/* tell the control that we will be reply asynchronously */
+ state->c = talloc_steal(state, c);
*async_reply = true;
return 0;
}
ctdb_disable_monitoring(ctdb);
- ret = ctdb_event_script_callback(ctdb,
- timeval_current_ofs(ctdb->tunable.script_timeout, 0),
- state,
+ ret = ctdb_event_script_callback(ctdb, state,
ctdb_start_recovery_callback,
- state, "startrecovery");
+ state, false,
+ CTDB_EVENT_START_RECOVERY,
+ "%s", "");
if (ret != 0) {
DEBUG(DEBUG_ERR,(__location__ " Failed to start recovery\n"));
return;
}
- DEBUG(DEBUG_ERR, ("Final timeout for recovery daemon ping. Shutting down ctdb daemon. (This can be caused if the cluster filesystem has hung)\n"));
+ DEBUG(DEBUG_ERR, ("Final timeout for recovery daemon ping. Restarting recovery daemon. (This can be caused if the cluster filesystem has hung)\n"));
ctdb_stop_recoverd(ctdb);
- ctdb_stop_keepalive(ctdb);
- ctdb_stop_monitoring(ctdb);
- ctdb_release_all_ips(ctdb);
- if (ctdb->methods != NULL) {
- ctdb->methods->shutdown(ctdb);
- }
- ctdb_event_script(ctdb, "shutdown");
- DEBUG(DEBUG_ERR, ("Recovery daemon ping timeout. Daemon has been shut down.\n"));
- exit(0);
+ ctdb_start_recoverd(ctdb);
}
/* The recovery daemon will ping us at regular intervals.
ctdb->recovery_master = ((uint32_t *)(&indata.dptr[0]))[0];
return 0;
}
+
+
+struct stop_node_callback_state {
+ struct ctdb_req_control *c;
+};
+
+/*
+ called when the 'stopped' event script has finished
+ */
+static void ctdb_stop_node_callback(struct ctdb_context *ctdb, int status, void *p)
+{
+ struct stop_node_callback_state *state = talloc_get_type(p, struct stop_node_callback_state);
+
+ if (status != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " stopped event script failed (status %d)\n", status));
+ ctdb->nodes[ctdb->pnn]->flags &= ~NODE_FLAGS_STOPPED;
+ if (status == -ETIME) {
+ ctdb_ban_self(ctdb);
+ }
+ }
+
+ ctdb_request_control_reply(ctdb, state->c, NULL, status, NULL);
+ talloc_free(state);
+}
+
+int32_t ctdb_control_stop_node(struct ctdb_context *ctdb, struct ctdb_req_control *c, bool *async_reply)
+{
+ int ret;
+ struct stop_node_callback_state *state;
+
+ DEBUG(DEBUG_INFO,(__location__ " Stopping node\n"));
+
+ state = talloc(ctdb, struct stop_node_callback_state);
+ CTDB_NO_MEMORY(ctdb, state);
+
+ state->c = talloc_steal(state, c);
+
+ ctdb_disable_monitoring(ctdb);
+
+ ret = ctdb_event_script_callback(ctdb, state,
+ ctdb_stop_node_callback,
+ state, false,
+ CTDB_EVENT_STOPPED, "%s", "");
+
+ if (ret != 0) {
+ ctdb_enable_monitoring(ctdb);
+
+ DEBUG(DEBUG_ERR,(__location__ " Failed to stop node\n"));
+ talloc_free(state);
+ return -1;
+ }
+
+ ctdb->nodes[ctdb->pnn]->flags |= NODE_FLAGS_STOPPED;
+
+ *async_reply = true;
+
+ return 0;
+}
+
+int32_t ctdb_control_continue_node(struct ctdb_context *ctdb)
+{
+ DEBUG(DEBUG_INFO,(__location__ " Continue node\n"));
+ ctdb->nodes[ctdb->pnn]->flags &= ~NODE_FLAGS_STOPPED;
+
+ return 0;
+}
git.samba.org / sahlberg / ctdb.git / blobdiff