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/network.h"
#include "system/filesys.h"
#include "lib/util/dlinklist.h"
#include "db_wrap.h"
+static bool later_db(const char *name)
+{
+ return (strstr(name, "notify") || strstr(name, "serverid"));
+}
/*
lock all databases
static int ctdb_lock_all_databases(struct ctdb_context *ctdb, uint32_t priority)
{
struct ctdb_db_context *ctdb_db;
+ /* REMOVE later */
+ /* This double loop is for backward compatibility and deadlock
+ avoidance for old samba versions that not yet support
+ the set prio call.
+ This code shall be removed later
+ */
for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
if (ctdb_db->priority != priority) {
continue;
}
+ if (later_db(ctdb_db->db_name)) {
+ continue;
+ }
DEBUG(DEBUG_INFO,("locking database 0x%08x priority:%u %s\n", ctdb_db->db_id, ctdb_db->priority, ctdb_db->db_name));
if (tdb_lockall(ctdb_db->ltdb->tdb) != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to lock database %s\n", ctdb_db->db_name));
+ return -1;
+ }
+ }
+ for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ if (ctdb_db->priority != priority) {
+ continue;
+ }
+ if (!later_db(ctdb_db->db_name)) {
+ continue;
+ }
+ DEBUG(DEBUG_INFO,("locking database 0x%08x priority:%u %s\n", ctdb_db->db_id, ctdb_db->priority, ctdb_db->db_name));
+ if (tdb_lockall(ctdb_db->ltdb->tdb) != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to lock database %s\n", ctdb_db->db_name));
return -1;
}
}
h->ctdb->freeze_mode[h->priority] = CTDB_FREEZE_FROZEN;
/* notify the waiters */
- while ((w = h->ctdb->freeze_handles[h->priority]->waiters)) {
+ if (h != h->ctdb->freeze_handles[h->priority]) {
+ DEBUG(DEBUG_ERR,("lockwait finished but h is not linked\n"));
+ }
+ while ((w = h->waiters)) {
w->status = status;
- DLIST_REMOVE(h->ctdb->freeze_handles[h->priority]->waiters, w);
+ DLIST_REMOVE(h->waiters, w);
talloc_free(w);
}
}
h->ctdb = ctdb;
h->priority = priority;
- /* use socketpair() instead of pipe() so we have bi-directional fds */
- if (socketpair(AF_UNIX, SOCK_STREAM, 0, fd) != 0) {
+ if (pipe(fd) == -1) {
DEBUG(DEBUG_ERR,("Failed to create pipe for ctdb_freeze_lock\n"));
talloc_free(h);
return NULL;
if (h->child == 0) {
int ret;
- int count = 0;
+
/* in the child */
close(fd[0]);
+
+ debug_extra = talloc_asprintf(NULL, "freeze_lock-%u:", priority);
ret = ctdb_lock_all_databases(ctdb, priority);
if (ret != 0) {
_exit(0);
}
- alarm(30);
-
- while (count++ < 30) {
- ret = write(fd[1], &ret, sizeof(ret));
- if (ret == sizeof(ret)) {
- break;
- }
+ ret = write(fd[1], &ret, sizeof(ret));
+ if (ret != sizeof(ret)) {
DEBUG(DEBUG_ERR, (__location__ " Failed to write to socket from freeze child. ret:%d errno:%u\n", ret, errno));
- sleep (1);
- }
- if (count >= 30) {
- DEBUG(DEBUG_ERR, (__location__ " Failed to write to socket from freeze child. Aborting freeze child\n"));
- _exit(0);
+ _exit(1);
}
- /* the read here means we will die if the parent exits */
- read(fd[1], &ret, sizeof(ret));
- _exit(0);
+ while (1) {
+ sleep(1);
+ if (kill(ctdb->ctdbd_pid, 0) != 0) {
+ DEBUG(DEBUG_ERR,("Parent died. Exiting lock wait child\n"));
+
+ _exit(0);
+ }
+ }
}
talloc_set_destructor(h, ctdb_freeze_handle_destructor);
close(fd[1]);
+ set_close_on_exec(fd[0]);
h->fd = fd[0];
- fde = event_add_fd(ctdb->ev, h, h->fd, EVENT_FD_READ|EVENT_FD_AUTOCLOSE,
+
+ fde = event_add_fd(ctdb->ev, h, h->fd, EVENT_FD_READ,
ctdb_freeze_lock_handler, h);
if (fde == NULL) {
DEBUG(DEBUG_ERR,("Failed to setup fd event for ctdb_freeze_lock\n"));
talloc_free(h);
return NULL;
}
+ tevent_fd_set_auto_close(fde);
return h;
}
*/
static int ctdb_freeze_waiter_destructor(struct ctdb_freeze_waiter *w)
{
- DLIST_REMOVE(w->ctdb->freeze_handles[w->priority]->waiters, w);
ctdb_request_control_reply(w->ctdb, w->c, NULL, w->status, NULL);
return 0;
}
*/
int ctdb_start_freeze(struct ctdb_context *ctdb, uint32_t priority)
{
+ if (priority == 0) {
+ DEBUG(DEBUG_ERR,("Freeze priority 0 requested, remapping to priority 1\n"));
+ priority = 1;
+ }
+
if ((priority < 1) || (priority > NUM_DB_PRIORITIES)) {
DEBUG(DEBUG_ERR,(__location__ " Invalid db priority : %u\n", priority));
return -1;
return 0;
}
+ /* Stop any vacuuming going on: we don't want to wait. */
+ ctdb_stop_vacuuming(ctdb);
+
/* if there isn't a freeze lock child then create one */
- if (!ctdb->freeze_handles[priority]) {
+ if (ctdb->freeze_handles[priority] == NULL) {
ctdb->freeze_handles[priority] = ctdb_freeze_lock(ctdb, priority);
CTDB_NO_MEMORY(ctdb, ctdb->freeze_handles[priority]);
ctdb->freeze_mode[priority] = CTDB_FREEZE_PENDING;
DEBUG(DEBUG_ERR, ("Freeze priority %u\n", priority));
+ if (priority == 0) {
+ DEBUG(DEBUG_ERR,("Freeze priority 0 requested, remapping to priority 1\n"));
+ priority = 1;
+ }
+
if ((priority < 1) || (priority > NUM_DB_PRIORITIES)) {
DEBUG(DEBUG_ERR,(__location__ " Invalid db priority : %u\n", priority));
return -1;
}
/* add ourselves to list of waiters */
+ if (ctdb->freeze_handles[priority] == NULL) {
+ DEBUG(DEBUG_ERR,("No freeze lock handle when adding a waiter\n"));
+ return -1;
+ }
+
w = talloc(ctdb->freeze_handles[priority], struct ctdb_freeze_waiter);
CTDB_NO_MEMORY(ctdb, w);
w->ctdb = ctdb;
system("mkdir -p test.db.saved; /usr/bin/rsync --delete -a test.db/ test.db.saved/$$ 2>&1 > /dev/null");
#endif
- talloc_free(ctdb->freeze_handles[priority]);
- ctdb->freeze_handles[priority] = NULL;
+ if (ctdb->freeze_handles[priority] != NULL) {
+ talloc_free(ctdb->freeze_handles[priority]);
+ ctdb->freeze_handles[priority] = NULL;
+ }
}
/*
return 0;
}
+/*
+ cancel a transaction for all databases - used for recovery
+ */
+int32_t ctdb_control_transaction_cancel(struct ctdb_context *ctdb)
+{
+ struct ctdb_db_context *ctdb_db;
+
+ DEBUG(DEBUG_ERR,(__location__ " recovery transaction cancelled called\n"));
+
+ for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ tdb_add_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
+
+ if (tdb_transaction_cancel(ctdb_db->ltdb->tdb) != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to cancel transaction for db '%s'\n", ctdb_db->db_name));
+ /* not a fatal error */
+ }
+
+ tdb_remove_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
+ }
+
+ ctdb->freeze_transaction_started = false;
+
+ return 0;
+}
+
/*
commit transactions on all databases
*/
{
struct ctdb_db_context *ctdb_db;
int i;
-
+ int healthy_nodes = 0;
+
for (i=1;i<=NUM_DB_PRIORITIES; i++) {
if (ctdb->freeze_mode[i] != CTDB_FREEZE_FROZEN) {
DEBUG(DEBUG_ERR,(__location__ " Failed transaction_start while not frozen\n"));
return -1;
}
+ DEBUG(DEBUG_DEBUG,(__location__ " num_nodes[%d]\n", ctdb->num_nodes));
+ for (i=0; i < ctdb->num_nodes; i++) {
+ DEBUG(DEBUG_DEBUG,(__location__ " node[%d].flags[0x%X]\n",
+ i, ctdb->nodes[i]->flags));
+ if (ctdb->nodes[i]->flags == 0) {
+ healthy_nodes++;
+ }
+ }
+ DEBUG(DEBUG_INFO,(__location__ " healthy_nodes[%d]\n", healthy_nodes));
+
for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ int ret;
+
tdb_add_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
- if (tdb_transaction_commit(ctdb_db->ltdb->tdb) != 0) {
+ ret = tdb_transaction_commit(ctdb_db->ltdb->tdb);
+ if (ret != 0) {
DEBUG(DEBUG_ERR,(__location__ " Failed to commit transaction for db '%s'. Cancel all transactions and resetting transaction_started to false.\n",
ctdb_db->db_name));
-
- /* cancel any pending transactions */
- for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
- tdb_add_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
- if (tdb_transaction_cancel(ctdb_db->ltdb->tdb) != 0) {
- DEBUG(DEBUG_ERR,(__location__ " Failed to cancel transaction for db '%s'\n",
- ctdb_db->db_name));
- }
- tdb_remove_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
- }
- ctdb->freeze_transaction_started = false;
-
- return -1;
+ goto fail;
}
tdb_remove_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
+
+ ret = ctdb_update_persistent_health(ctdb, ctdb_db, NULL, healthy_nodes);
+ if (ret != 0) {
+ DEBUG(DEBUG_CRIT,(__location__ " Failed to update persistent health for db '%s'. "
+ "Cancel all remaining transactions and resetting transaction_started to false.\n",
+ ctdb_db->db_name));
+ goto fail;
+ }
}
ctdb->freeze_transaction_started = false;
ctdb->freeze_transaction_id = 0;
return 0;
+
+fail:
+ /* cancel any pending transactions */
+ for (ctdb_db=ctdb->db_list;ctdb_db;ctdb_db=ctdb_db->next) {
+ tdb_add_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
+ if (tdb_transaction_cancel(ctdb_db->ltdb->tdb) != 0) {
+ DEBUG(DEBUG_ERR,(__location__ " Failed to cancel transaction for db '%s'\n",
+ ctdb_db->db_name));
+ }
+ tdb_remove_flags(ctdb_db->ltdb->tdb, TDB_NOLOCK);
+ }
+ ctdb->freeze_transaction_started = false;
+
+ return -1;
}
/*
git.samba.org / sahlberg / ctdb.git / blobdiff