bool Connection::ping()
{
log_debug("ping()");
if (PQsendQuery(conn, "select 1") == 0)
{
log_debug("failed to send statement \"select 1\" to database in Connection::ping()");
return false;
}
while (true)
{
struct pollfd fd;
fd.fd = PQsocket(conn);
fd.events = POLLIN;
log_debug("wait for input on fd " << fd.fd);
if (::poll(&fd, 1, 10000) != 1)
{
log_debug("no data received in Connection::ping()");
return false;
}
log_debug("consumeInput");
if (PQconsumeInput(conn) == 0)
{
log_debug("PQconsumeInput failed in Connection::ping()");
return false;
}
log_debug("check PQisBusy");
while (PQisBusy(conn) == 0)
{
log_debug("PQgetResult");
PGresult* result = PQgetResult(conn);
log_debug("PQgetResult => " << static_cast<void*>(result));
if (result == 0)
return true;
log_debug("PQfree");
PQclear(result);
}
}
}
PGconn * pgpool_getfreeconn(void)
{
assert(initialized);
int id = last_used;
while (1) {
id = (id+1) % poolsize;
PGconn *c = pool[id];
assert(c);
PGresult *res;
int rc;
// can we get state AND is the server
// not blocking and any results?
while ((rc=PQconsumeInput(c)) && !PQisBusy(c)
&& (res=PQgetResult(c))) {
// we have results. we need to clear those resultsets
if (reshandler(c, res)) {
return NULL;
}
}
if (!rc) {
log_fatal("pgpool", "Error in PQconsumeInput. %s",
PQerrorMessage(c));
return NULL;
} else if (!PQisBusy(c)) {
last_used = id;
return c;
}
if (id == last_used) {
usleep(150000);
}
}
}
/* Checks whether new data has arrived from the server (on either the snapshot
* connection or the replication connection, as appropriate). If yes, it is
* processed, and context->status is set to 1. If no data is available, this
* function does not block, but returns immediately, and context->status is set
* to 0. If the data stream has ended, context->status is set to -1. */
int db_client_poll(client_context_t context) {
int err = 0;
if (context->sql_conn) {
/* To make PQgetResult() non-blocking, check PQisBusy() first */
if (PQisBusy(context->sql_conn)) {
context->status = 0;
return err;
}
check(err, snapshot_poll(context));
context->status = 1;
/* If the snapshot is finished, switch over to the replication stream */
if (!context->sql_conn) {
checkRepl(err, context, replication_stream_start(&context->repl));
}
return err;
} else {
checkRepl(err, context, replication_stream_poll(&context->repl));
context->status = context->repl.status;
return err;
}
}
static void pq_event(evutil_socket_t fd, short event, void *arg) {
struct connection_struct* database = (struct connection_struct*) arg;
if (database->queries) {
if (database->queries->sent == 0) {
PQsendQuery(database->conn, database->queries->query);
database->queries->sent = 1;
}
if (PQconsumeInput(database->conn) && !PQisBusy(database->conn)) {
PGresult* res = PQgetResult(database->conn);
while (res) {
if (database->queries->callback)
database->queries->callback(res, database->queries->context, database->queries->query);
if (database->report_errors && PQresultStatus(res) != PGRES_COMMAND_OK)
fprintf(stderr, "Query: '%s' returned error\n\t%s\n", database->queries->query, PQresultErrorMessage(res));
PQclear(res);
res = PQgetResult(database->conn);
}
database->query_count--;
struct query_struct* old = database->queries;
database->queries = database->queries->next;
free(old->query);
free(old);
pq_event(fd, event, arg);
}
}
}
static void
evpg_query_finished(int sock, short which, void **data)
{
struct evpg_db_node *dbnode;
struct evpg_cfg *config;
const char *querystr;
void (*cb)(PGresult *, void *);
void *usrdata;
struct event *event;
config = data[0];
querystr = data[1];
cb = data[2];
usrdata = data[3];
dbnode = data[4];
event = data[5];
PQconsumeInput(dbnode->dbconn);
if (PQisBusy(dbnode->dbconn) == 0)
{
PGresult *result;
result = PQgetResult(dbnode->dbconn);
cb(result, usrdata);
PQclear(result);
free(event);
free(data);
evpg_set_ready(config, dbnode);
return;
}
/* this query has not finished */
event_set(event, sock, EV_READ, (void *)evpg_query_finished, data);
event_add(event, 0);
}
/*
* Send a query and wait for the results by using the asynchronous libpq
* functions and the backend version of select().
*
* We must not use the regular blocking libpq functions like PQexec()
* since they are uninterruptible by signals on some platforms, such as
* Windows.
*
* We must also not use vanilla select() here since it cannot handle the
* signal emulation layer on Windows.
*
* The function is modeled on PQexec() in libpq, but only implements
* those parts that are in use in the walreceiver.
*
* Queries are always executed on the connection in streamConn.
*/
static PGresult *
libpqrcv_PQexec(const char *query)
{
PGresult *result = NULL;
PGresult *lastResult = NULL;
/*
* PQexec() silently discards any prior query results on the connection.
* This is not required for walreceiver since it's expected that walsender
* won't generate any such junk results.
*/
/*
* Submit a query. Since we don't use non-blocking mode, this also can
* block. But its risk is relatively small, so we ignore that for now.
*/
if (!PQsendQuery(streamConn, query))
return NULL;
for (;;)
{
/*
* Receive data until PQgetResult is ready to get the result without
* blocking.
*/
while (PQisBusy(streamConn))
{
/*
* We don't need to break down the sleep into smaller increments,
* and check for interrupts after each nap, since we can just
* elog(FATAL) within SIGTERM signal handler if the signal arrives
* in the middle of establishment of replication connection.
*/
if (!libpq_select(-1))
continue; /* interrupted */
if (PQconsumeInput(streamConn) == 0)
return NULL; /* trouble */
}
/*
* Emulate the PQexec()'s behavior of returning the last result when
* there are many. Since walsender will never generate multiple
* results, we skip the concatenation of error messages.
*/
result = PQgetResult(streamConn);
if (result == NULL)
break; /* query is complete */
PQclear(lastResult);
lastResult = result;
if (PQresultStatus(lastResult) == PGRES_COPY_IN ||
PQresultStatus(lastResult) == PGRES_COPY_OUT ||
PQresultStatus(lastResult) == PGRES_COPY_BOTH ||
PQstatus(streamConn) == CONNECTION_BAD)
break;
}
return lastResult;
}
PGresult * do_postgres_cCommand_execute_async(VALUE self, VALUE connection, PGconn *db, VALUE query) {
PGresult *response;
char* str = StringValuePtr(query);
while ((response = PQgetResult(db))) {
PQclear(response);
}
struct timeval start;
int retval;
gettimeofday(&start, NULL);
retval = PQsendQuery(db, str);
if (!retval) {
if (PQstatus(db) != CONNECTION_OK) {
PQreset(db);
if (PQstatus(db) == CONNECTION_OK) {
retval = PQsendQuery(db, str);
}
else {
do_postgres_full_connect(connection, db);
retval = PQsendQuery(db, str);
}
}
if (!retval) {
rb_raise(eDO_ConnectionError, "%s", PQerrorMessage(db));
}
}
int socket_fd = PQsocket(db);
fd_set rset;
while (1) {
FD_ZERO(&rset);
FD_SET(socket_fd, &rset);
retval = rb_thread_select(socket_fd + 1, &rset, NULL, NULL, NULL);
if (retval < 0) {
rb_sys_fail(0);
}
if (retval == 0) {
continue;
}
if (PQconsumeInput(db) == 0) {
rb_raise(eDO_ConnectionError, "%s", PQerrorMessage(db));
}
if (PQisBusy(db) == 0) {
break;
}
}
data_objects_debug(connection, query, &start);
return PQgetResult(db);
}
static int
esql_postgresql_io(Esql *e)
{
if (PQstatus(e->backend.db) == CONNECTION_BAD)
{
ERR("%s", esql_postgresql_error_get(e));
return ECORE_FD_ERROR;
}
if (e->current == ESQL_CONNECT_TYPE_INIT)
{
switch (PQconnectPoll(e->backend.db))
{
case PGRES_POLLING_OK:
return 0;
case PGRES_POLLING_READING:
return ECORE_FD_READ;
case PGRES_POLLING_WRITING:
return ECORE_FD_WRITE;
default:
ERR("%s", esql_postgresql_error_get(e));
return ECORE_FD_ERROR;
}
}
if (!PQconsumeInput(e->backend.db))
{
ERR("%s", esql_postgresql_error_get(e));
return ECORE_FD_ERROR;
}
if (!PQisBusy(e->backend.db)) return 0;
return ECORE_FD_READ | ECORE_FD_WRITE; /* psql does not provide a method to get read/write mode :( */
}
/*
* wait until current query finishes ignoring any results, this could be an async command
* or a cancelation of a query
* return 1 if Ok; 0 if any error ocurred; -1 if timeout reached
*/
int
wait_connection_availability(PGconn *conn, int timeout)
{
PGresult *res;
while(timeout-- >= 0)
{
if (PQconsumeInput(conn) == 0)
{
log_warning(_("PQconsumeInput: Query could not be sent to primary. %s\n"),
PQerrorMessage(conn));
return 0;
}
if (PQisBusy(conn) == 0)
{
res = PQgetResult(conn);
if (res == NULL)
break;
PQclear(res);
}
sleep(1);
}
if (timeout >= 0)
return 1;
else
return -1;
}
static void
write_queue(ParallelWriter *self, const void *buffer, uint32 len)
{
struct iovec iov[2];
AssertArg(self->conn != NULL);
AssertArg(self->queue != NULL);
AssertArg(len == 0 || buffer != NULL);
iov[0].iov_base = &len;
iov[0].iov_len = sizeof(len);
iov[1].iov_base = (void *) buffer;
iov[1].iov_len = len;
for (;;)
{
if (QueueWrite(self->queue, iov, 2, DEFAULT_TIMEOUT_MSEC, false))
return;
PQconsumeInput(self->conn);
if (!PQisBusy(self->conn))
{
ereport(ERROR,
(errcode(ERRCODE_SQLCLIENT_UNABLE_TO_ESTABLISH_SQLCONNECTION),
errmsg("unexpected reader termination"),
errdetail("%s", finish_and_get_message(self))));
}
/* retry */
}
}
/*
* Helper function that actually kicks off the command on the libpq connection.
*/
static void
dispatchCommand(CdbDispatchResult * dispatchResult,
const char *query_text,
int query_text_len)
{
SegmentDatabaseDescriptor *segdbDesc = dispatchResult->segdbDesc;
TimestampTz beforeSend = 0;
long secs;
int usecs;
if (DEBUG1 >= log_min_messages)
beforeSend = GetCurrentTimestamp();
if (PQisBusy(segdbDesc->conn))
elog(LOG, "Trying to send to busy connection %s: asyncStatus %d",
segdbDesc->whoami,
segdbDesc->conn->asyncStatus);
if (cdbconn_isBadConnection(segdbDesc))
{
char *msg = PQerrorMessage(dispatchResult->segdbDesc->conn);
dispatchResult->stillRunning = false;
ereport(ERROR,
(errcode(ERRCODE_GP_INTERCONNECTION_ERROR),
errmsg("Connection lost before dispatch to segment %s: %s",
dispatchResult->segdbDesc->whoami, msg ? msg : "unknown error")));
}
/*
* Submit the command asynchronously.
*/
if (PQsendGpQuery_shared(dispatchResult->segdbDesc->conn, (char *) query_text, query_text_len) == 0)
{
char *msg = PQerrorMessage(dispatchResult->segdbDesc->conn);
dispatchResult->stillRunning = false;
ereport(ERROR,
(errcode(ERRCODE_GP_INTERCONNECTION_ERROR),
errmsg("Command could not be dispatch to segment %s: %s",
dispatchResult->segdbDesc->whoami, msg ? msg : "unknown error")));
}
if (DEBUG1 >= log_min_messages)
{
TimestampDifference(beforeSend, GetCurrentTimestamp(), &secs, &usecs);
if (secs != 0 || usecs > 1000) /* Time > 1ms? */
elog(LOG, "time for PQsendGpQuery_shared %ld.%06d", secs, usecs);
}
/*
* We'll keep monitoring this QE -- whether or not the command
* was dispatched -- in order to check for a lost connection
* or any other errors that libpq might have in store for us.
*/
dispatchResult->stillRunning = true;
dispatchResult->hasDispatched = true;
ELOG_DISPATCHER_DEBUG("Command dispatched to QE (%s)", dispatchResult->segdbDesc->whoami);
}
int
pgut_wait(int num, PGconn *connections[], struct timeval *timeout)
{
/* all connections are busy. wait for finish */
while (!interrupted)
{
int i;
fd_set mask;
int maxsock;
FD_ZERO(&mask);
maxsock = -1;
for (i = 0; i < num; i++)
{
int sock;
if (connections[i] == NULL)
continue;
sock = PQsocket(connections[i]);
if (sock >= 0)
{
FD_SET(sock, &mask);
if (maxsock < sock)
maxsock = sock;
}
}
if (maxsock == -1)
{
errno = ENOENT;
return -1;
}
i = wait_for_sockets(maxsock + 1, &mask, timeout);
if (i == 0)
break; /* timeout */
for (i = 0; i < num; i++)
{
if (connections[i] && FD_ISSET(PQsocket(connections[i]), &mask))
{
PQconsumeInput(connections[i]);
if (PQisBusy(connections[i]))
continue;
return i;
}
}
}
errno = EINTR;
return -1;
}
int db_postgres_async_resume(db_con_t *_h, int fd, db_res_t **_r, void *_priv)
{
struct pool_con *con = (struct pool_con *)_priv;
PGresult *res = NULL;
#ifdef EXTRA_DEBUG
if (!db_match_async_con(fd, _h)) {
LM_BUG("no conn match for fd %d", fd);
abort();
}
#endif
db_switch_to_async(_h, con);
if( PQconsumeInput(CON_CONNECTION(_h)) == 0) {
LM_ERR("Unable to consume input\n");
db_switch_to_sync(_h);
db_store_async_con(_h, con);
return -1;
}
if(PQisBusy(CON_CONNECTION(_h))) {
async_status = ASYNC_CONTINUE;
db_switch_to_sync(_h);
return 1;
}
while (1) {
if ((res = PQgetResult(CON_CONNECTION(_h)))) {
CON_RESULT(_h) = res;
} else {
break;
}
}
if (_r) {
if (db_postgres_store_result(_h, _r) != 0) {
LM_ERR("failed to store result\n");
db_switch_to_sync(_h);
db_store_async_con(_h, con);
return -2;
}
}
db_switch_to_sync(_h);
db_store_async_con(_h, con);
return 0;
}
static void
close_connections()
{
if (primary_conn != NULL && PQisBusy(primary_conn) == 1)
cancel_query(primary_conn, local_options.master_response_timeout);
if (my_local_conn != NULL)
PQfinish(my_local_conn);
if (primary_conn != NULL && primary_conn != my_local_conn)
PQfinish(primary_conn);
primary_conn = NULL;
my_local_conn = NULL;
}
/*
* call-seq:
* conn.fetch() -> result or nil
* conn.fetch() { |result| ... } -> obj
*
* Fetches the results of the previous Pg::Conn#send call.
* See there for an example.
*
* The result will be +nil+ if there are no more results.
*/
VALUE
pgconn_fetch( VALUE self)
{
struct pgconn_data *c;
PGresult *result;
VALUE res;
Data_Get_Struct( self, struct pgconn_data, c);
pg_check_conninvalid( c);
if (PQconsumeInput( c->conn) == 0)
pg_raise_connexec( c);
if (PQisBusy( c->conn) > 0)
return Qnil;
result = PQgetResult( c->conn);
return result == NULL ? Qnil :
yield_or_return_result( pgresult_new( result, c, Qnil, Qnil));
}
ngx_int_t
ngx_postgres_upstream_get_ack(ngx_http_request_t *r, ngx_connection_t *pgxc,
ngx_postgres_upstream_peer_data_t *pgdt)
{
PGresult *res;
dd("entering");
if (!PQconsumeInput(pgdt->pgconn)) {
dd("returning NGX_ERROR");
return NGX_ERROR;
}
if (PQisBusy(pgdt->pgconn)) {
dd("returning NGX_AGAIN");
return NGX_AGAIN;
}
/* remove result timeout */
if (pgxc->read->timer_set) {
ngx_del_timer(pgxc->read);
}
dd("receiving ACK (ready for next query)");
res = PQgetResult(pgdt->pgconn);
if (res != NULL) {
dd("receiving ACK failed");
ngx_log_error(NGX_LOG_ERR, pgxc->log, 0,
"receiving ACK failed: multiple queries(?)");
PQclear(res);
dd("returning NGX_HTTP_INTERNAL_SERVER_ERROR");
return NGX_HTTP_INTERNAL_SERVER_ERROR;
}
dd("ACK received successfully");
pgxc->log->action = "being idle on PostgreSQL database";
pgdt->state = state_db_idle;
dd("returning");
return ngx_postgres_upstream_done(r, r->upstream, pgdt);
}
void
ngx_postgres_keepalive_close_handler(ngx_event_t *ev)
{
ngx_postgres_upstream_srv_conf_t *pgscf;
ngx_postgres_keepalive_cache_t *item;
ngx_connection_t *c;
PGresult *res;
dd("entering");
c = ev->data;
item = c->data;
if (c->close) {
goto close;
}
if (PQconsumeInput(item->pgconn) && !PQisBusy(item->pgconn)) {
res = PQgetResult(item->pgconn);
if (res == NULL) {
dd("returning");
return;
}
PQclear(res);
dd("received result on idle keepalive connection");
ngx_log_error(NGX_LOG_ERR, c->log, 0,
"postgres: received result on idle keepalive connection");
}
close:
pgscf = item->srv_conf;
ngx_postgres_upstream_free_connection(ev->log, c, item->pgconn, pgscf);
ngx_queue_remove(&item->queue);
ngx_queue_insert_head(&pgscf->free, &item->queue);
dd("returning");
}
/* MultiClientResultStatus checks result status for an asynchronous query. */
ResultStatus
MultiClientResultStatus(int32 connectionId)
{
PGconn *connection = NULL;
int consumed = 0;
ConnStatusType connStatusType = CONNECTION_OK;
ResultStatus resultStatus = CLIENT_INVALID_RESULT_STATUS;
Assert(connectionId != INVALID_CONNECTION_ID);
connection = ClientConnectionArray[connectionId];
Assert(connection != NULL);
connStatusType = PQstatus(connection);
if (connStatusType == CONNECTION_BAD)
{
ereport(WARNING, (errmsg("could not maintain connection to worker node")));
return CLIENT_RESULT_UNAVAILABLE;
}
/* consume input to allow status change */
consumed = PQconsumeInput(connection);
if (consumed != 0)
{
int connectionBusy = PQisBusy(connection);
if (connectionBusy == 0)
{
resultStatus = CLIENT_RESULT_READY;
}
else
{
resultStatus = CLIENT_RESULT_BUSY;
}
}
else
{
ereport(WARNING, (errmsg("could not consume data from worker node")));
resultStatus = CLIENT_RESULT_UNAVAILABLE;
}
return resultStatus;
}
/*
* Wait for the result from a prior asynchronous execution function call.
*
* This function offers quick responsiveness by checking for any interruptions.
*
* This function emulates the PQexec()'s behavior of returning the last result
* when there are many.
*
* Caller is responsible for the error handling on the result.
*/
PGresult *
pgfdw_get_result(PGconn *conn, const char *query)
{
PGresult *last_res = NULL;
for (;;)
{
PGresult *res;
while (PQisBusy(conn))
{
int wc;
/* Sleep until there's something to do */
wc = WaitLatchOrSocket(MyLatch,
WL_LATCH_SET | WL_SOCKET_READABLE,
PQsocket(conn),
-1L, PG_WAIT_EXTENSION);
ResetLatch(MyLatch);
CHECK_FOR_INTERRUPTS();
/* Data available in socket */
if (wc & WL_SOCKET_READABLE)
{
if (!PQconsumeInput(conn))
pgfdw_report_error(ERROR, NULL, conn, false, query);
}
}
res = PQgetResult(conn);
if (res == NULL)
break; /* query is complete */
PQclear(last_res);
last_res = res;
}
return last_res;
}
static void ready_input(ErlDrvData drv_data, ErlDrvEvent event)
{
our_data_t* data = (our_data_t*)drv_data;
PGconn* conn = data->conn;
ei_x_buff x;
PGresult* res;
PQconsumeInput(conn);
if (PQisBusy(conn))
return;
ei_x_new_with_version(&x);
res = PQgetResult(conn);
encode_result(&x, res, conn);
driver_output(data->port, x.buff, x.index);
ei_x_free(&x);
PQclear(res);
for (;;) {
res = PQgetResult(conn);
if (res == NULL)
break;
PQclear(res);
}
}
ngx_int_t
ngx_postgres_upstream_get_result(ngx_http_request_t *r, ngx_connection_t *pgxc,
ngx_postgres_upstream_peer_data_t *pgdt)
{
ExecStatusType pgrc;
PGresult *res;
ngx_int_t rc;
dd("entering");
/* remove connection timeout from re-used keepalive connection */
if (pgxc->write->timer_set) {
ngx_del_timer(pgxc->write);
}
if (!PQconsumeInput(pgdt->pgconn)) {
dd("returning NGX_ERROR");
return NGX_ERROR;
}
if (PQisBusy(pgdt->pgconn)) {
dd("returning NGX_AGAIN");
return NGX_AGAIN;
}
dd("receiving result");
res = PQgetResult(pgdt->pgconn);
if (res == NULL) {
dd("receiving result failed");
ngx_log_error(NGX_LOG_ERR, pgxc->log, 0,
"receiving result failed: %s",
PQerrorMessage(pgdt->pgconn));
dd("returning NGX_ERROR");
return NGX_ERROR;
}
pgrc = PQresultStatus(res);
if ((pgrc != PGRES_COMMAND_OK) && (pgrc != PGRES_TUPLES_OK)) {
dd("receiving result failed");
ngx_log_error(NGX_LOG_ERR, pgxc->log, 0,
"receiving result failed: %s: %s",
PQresStatus(pgrc),
PQerrorMessage(pgdt->pgconn));
PQclear(res);
dd("returning NGX_HTTP_INTERNAL_SERVER_ERROR");
return NGX_HTTP_INTERNAL_SERVER_ERROR;
}
dd("result received successfully, cols:%d rows:%d",
PQnfields(res), PQntuples(res));
pgxc->log->action = "processing result from PostgreSQL database";
rc = ngx_postgres_process_response(r, res);
PQclear(res);
if (rc != NGX_DONE) {
dd("returning rc:%d", (int) rc);
return rc;
}
dd("result processed successfully");
pgxc->log->action = "waiting for ACK from PostgreSQL database";
pgdt->state = state_db_get_ack;
dd("returning");
return ngx_postgres_upstream_get_ack(r, pgxc, pgdt);
}
/*
* Our caller already sent the query associated with this step. Wait for it
* to either complete or (if given the STEP_NONBLOCK flag) to block while
* waiting for a lock. We assume that any lock wait will persist until we
* have executed additional steps in the permutation.
*
* When calling this function on behalf of a given step for a second or later
* time, pass the STEP_RETRY flag. This only affects the messages printed.
*
* If the STEP_NONBLOCK flag was specified and the query is waiting to acquire
* a lock, returns true. Otherwise, returns false.
*/
static bool
try_complete_step(Step *step, int flags)
{
PGconn *conn = conns[1 + step->session];
fd_set read_set;
struct timeval timeout;
int sock = PQsocket(conn);
int ret;
PGresult *res;
FD_ZERO(&read_set);
while (flags & STEP_NONBLOCK && PQisBusy(conn))
{
FD_SET(sock, &read_set);
timeout.tv_sec = 0;
timeout.tv_usec = 10000; /* Check for lock waits every 10ms. */
ret = select(sock + 1, &read_set, NULL, NULL, &timeout);
if (ret < 0) /* error in select() */
{
fprintf(stderr, "select failed: %s\n", strerror(errno));
exit_nicely();
}
else if (ret == 0) /* select() timeout: check for lock wait */
{
int ntuples;
res = PQexecPrepared(conns[0], PREP_WAITING, 1,
&backend_pids[step->session + 1],
NULL, NULL, 0);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
fprintf(stderr, "lock wait query failed: %s",
PQerrorMessage(conn));
exit_nicely();
}
ntuples = PQntuples(res);
PQclear(res);
if (ntuples >= 1) /* waiting to acquire a lock */
{
if (!(flags & STEP_RETRY))
printf("step %s: %s <waiting ...>\n",
step->name, step->sql);
return true;
}
/* else, not waiting: give it more time */
}
else if (!PQconsumeInput(conn)) /* select(): data available */
{
fprintf(stderr, "PQconsumeInput failed: %s", PQerrorMessage(conn));
exit_nicely();
}
}
if (flags & STEP_RETRY)
printf("step %s: <... completed>\n", step->name);
else
printf("step %s: %s\n", step->name, step->sql);
while ((res = PQgetResult(conn)))
{
switch (PQresultStatus(res))
{
case PGRES_COMMAND_OK:
break;
case PGRES_TUPLES_OK:
printResultSet(res);
break;
case PGRES_FATAL_ERROR:
/* Detail may contain xid values, so just show primary. */
printf("%s: %s\n", PQresultErrorField(res, PG_DIAG_SEVERITY),
PQresultErrorField(res, PG_DIAG_MESSAGE_PRIMARY));
break;
default:
printf("unexpected result status: %s\n",
PQresStatus(PQresultStatus(res)));
}
PQclear(res);
}
return false;
}
/*
* GetIdleSlot
* Return a connection slot that is ready to execute a command.
*
* We return the first slot we find that is marked isFree, if one is;
* otherwise, we loop on select() until one socket becomes available. When
* this happens, we read the whole set and mark as free all sockets that become
* available.
*
* If an error occurs, NULL is returned.
*/
static ParallelSlot *
GetIdleSlot(ParallelSlot slots[], int numslots,
const char *progname)
{
int i;
int firstFree = -1;
/* Any connection already known free? */
for (i = 0; i < numslots; i++)
{
if (slots[i].isFree)
return slots + i;
}
/*
* No free slot found, so wait until one of the connections has finished
* its task and return the available slot.
*/
while (firstFree < 0)
{
fd_set slotset;
int maxFd = 0;
bool aborting;
/* We must reconstruct the fd_set for each call to select_loop */
FD_ZERO(&slotset);
for (i = 0; i < numslots; i++)
{
int sock = PQsocket(slots[i].connection);
/*
* We don't really expect any connections to lose their sockets
* after startup, but just in case, cope by ignoring them.
*/
if (sock < 0)
continue;
FD_SET(sock, &slotset);
if (sock > maxFd)
maxFd = sock;
}
SetCancelConn(slots->connection);
i = select_loop(maxFd, &slotset, &aborting);
ResetCancelConn();
if (aborting)
{
/*
* We set the cancel-receiving connection to the one in the zeroth
* slot above, so fetch the error from there.
*/
GetQueryResult(slots->connection, progname);
return NULL;
}
Assert(i != 0);
for (i = 0; i < numslots; i++)
{
int sock = PQsocket(slots[i].connection);
if (sock >= 0 && FD_ISSET(sock, &slotset))
{
/* select() says input is available, so consume it */
PQconsumeInput(slots[i].connection);
}
/* Collect result(s) as long as any are available */
while (!PQisBusy(slots[i].connection))
{
PGresult *result = PQgetResult(slots[i].connection);
if (result != NULL)
{
/* Check and discard the command result */
if (!ProcessQueryResult(slots[i].connection, result,
progname))
return NULL;
}
else
{
/* This connection has become idle */
slots[i].isFree = true;
if (firstFree < 0)
firstFree = i;
break;
}
}
}
}
return slots + firstFree;
}
/*
* pqSetenvPoll
*
* Polls the process of passing the values of a standard set of environment
* variables to the backend.
*/
PostgresPollingStatusType
pqSetenvPoll(PGconn *conn)
{
PGresult *res;
if (conn == NULL || conn->status == CONNECTION_BAD)
return PGRES_POLLING_FAILED;
/* Check whether there are any data for us */
switch (conn->setenv_state)
{
/* These are reading states */
case SETENV_STATE_CLIENT_ENCODING_WAIT:
case SETENV_STATE_OPTION_WAIT:
case SETENV_STATE_QUERY1_WAIT:
case SETENV_STATE_QUERY2_WAIT:
{
/* Load waiting data */
int n = pqReadData(conn);
if (n < 0)
goto error_return;
if (n == 0)
return PGRES_POLLING_READING;
break;
}
/* These are writing states, so we just proceed. */
case SETENV_STATE_CLIENT_ENCODING_SEND:
case SETENV_STATE_OPTION_SEND:
case SETENV_STATE_QUERY1_SEND:
case SETENV_STATE_QUERY2_SEND:
break;
/* Should we raise an error if called when not active? */
case SETENV_STATE_IDLE:
return PGRES_POLLING_OK;
default:
printfPQExpBuffer(&conn->errorMessage,
libpq_gettext(
"invalid setenv state %c, "
"probably indicative of memory corruption\n"
),
conn->setenv_state);
goto error_return;
}
/* We will loop here until there is nothing left to do in this call. */
for (;;)
{
switch (conn->setenv_state)
{
/*
* The _CLIENT_ENCODING_SEND code is slightly different from
* _OPTION_SEND below (e.g., no getenv() call), which is why a
* different state is used.
*/
case SETENV_STATE_CLIENT_ENCODING_SEND:
{
char setQuery[100]; /* note length limit in
* sprintf below */
const char *val = conn->client_encoding_initial;
if (val)
{
if (pg_strcasecmp(val, "default") == 0)
sprintf(setQuery, "SET client_encoding = DEFAULT");
else
sprintf(setQuery, "SET client_encoding = '%.60s'",
val);
#ifdef CONNECTDEBUG
fprintf(stderr,
"Sending client_encoding with %s\n",
setQuery);
#endif
if (!PQsendQuery(conn, setQuery))
goto error_return;
conn->setenv_state = SETENV_STATE_CLIENT_ENCODING_WAIT;
}
else
conn->setenv_state = SETENV_STATE_OPTION_SEND;
break;
}
case SETENV_STATE_OPTION_SEND:
{
/*
* Send SET commands for stuff directed by Environment
* Options. Note: we assume that SET commands won't start
* transaction blocks, even in a 7.3 server with
* autocommit off.
*/
char setQuery[100]; /* note length limit in
* sprintf below */
if (conn->next_eo->envName)
{
const char *val;
if ((val = getenv(conn->next_eo->envName)))
{
if (pg_strcasecmp(val, "default") == 0)
sprintf(setQuery, "SET %s = DEFAULT",
conn->next_eo->pgName);
else
sprintf(setQuery, "SET %s = '%.60s'",
conn->next_eo->pgName, val);
#ifdef CONNECTDEBUG
fprintf(stderr,
"Use environment variable %s to send %s\n",
conn->next_eo->envName, setQuery);
#endif
if (!PQsendQuery(conn, setQuery))
goto error_return;
conn->setenv_state = SETENV_STATE_OPTION_WAIT;
}
else
conn->next_eo++;
}
else
{
/* No more options to send, so move on to querying */
conn->setenv_state = SETENV_STATE_QUERY1_SEND;
}
break;
}
case SETENV_STATE_CLIENT_ENCODING_WAIT:
{
if (PQisBusy(conn))
return PGRES_POLLING_READING;
res = PQgetResult(conn);
if (res)
{
if (PQresultStatus(res) != PGRES_COMMAND_OK)
{
PQclear(res);
goto error_return;
}
PQclear(res);
/* Keep reading until PQgetResult returns NULL */
}
else
{
/* Query finished, so send the next option */
conn->setenv_state = SETENV_STATE_OPTION_SEND;
}
break;
}
case SETENV_STATE_OPTION_WAIT:
{
if (PQisBusy(conn))
return PGRES_POLLING_READING;
res = PQgetResult(conn);
if (res)
{
if (PQresultStatus(res) != PGRES_COMMAND_OK)
{
PQclear(res);
goto error_return;
}
PQclear(res);
/* Keep reading until PQgetResult returns NULL */
}
else
{
/* Query finished, so send the next option */
conn->next_eo++;
conn->setenv_state = SETENV_STATE_OPTION_SEND;
}
break;
}
case SETENV_STATE_QUERY1_SEND:
{
/*
* Issue query to get information we need. Here we must
* use begin/commit in case autocommit is off by default
* in a 7.3 server.
*
* Note: version() exists in all protocol-2.0-supporting
* backends. In 7.3 it would be safer to write
* pg_catalog.version(), but we can't do that without
* causing problems on older versions.
*/
if (!PQsendQuery(conn, "begin; select version(); end"))
goto error_return;
conn->setenv_state = SETENV_STATE_QUERY1_WAIT;
return PGRES_POLLING_READING;
}
case SETENV_STATE_QUERY1_WAIT:
{
if (PQisBusy(conn))
return PGRES_POLLING_READING;
res = PQgetResult(conn);
if (res)
{
char *val;
if (PQresultStatus(res) == PGRES_COMMAND_OK)
{
/* ignore begin/commit command results */
PQclear(res);
continue;
}
if (PQresultStatus(res) != PGRES_TUPLES_OK ||
PQntuples(res) != 1)
{
PQclear(res);
goto error_return;
}
/*
* Extract server version and save as if
* ParameterStatus
*/
val = PQgetvalue(res, 0, 0);
if (val && strncmp(val, "PostgreSQL ", 11) == 0)
{
char *ptr;
/* strip off PostgreSQL part */
val += 11;
/*
* strip off platform part (scribbles on result,
* naughty naughty)
*/
ptr = strchr(val, ' ');
if (ptr)
*ptr = '\0';
pqSaveParameterStatus(conn, "server_version",
val);
}
PQclear(res);
/* Keep reading until PQgetResult returns NULL */
}
else
{
/* Query finished, move to next */
conn->setenv_state = SETENV_STATE_QUERY2_SEND;
}
break;
}
case SETENV_STATE_QUERY2_SEND:
{
const char *query;
/*
* pg_client_encoding does not exist in pre-7.2 servers.
* So we need to be prepared for an error here. Do *not*
* start a transaction block, except in 7.3 servers where
* we need to prevent autocommit-off from starting a
* transaction anyway.
*/
if (conn->sversion >= 70300 &&
conn->sversion < 70400)
query = "begin; select pg_catalog.pg_client_encoding(); end";
else
query = "select pg_client_encoding()";
if (!PQsendQuery(conn, query))
goto error_return;
conn->setenv_state = SETENV_STATE_QUERY2_WAIT;
return PGRES_POLLING_READING;
}
case SETENV_STATE_QUERY2_WAIT:
{
if (PQisBusy(conn))
return PGRES_POLLING_READING;
res = PQgetResult(conn);
if (res)
{
const char *val;
if (PQresultStatus(res) == PGRES_COMMAND_OK)
{
/* ignore begin/commit command results */
PQclear(res);
continue;
}
if (PQresultStatus(res) == PGRES_TUPLES_OK &&
PQntuples(res) == 1)
{
/* Extract client encoding and save it */
val = PQgetvalue(res, 0, 0);
if (val && *val) /* null should not happen, but */
pqSaveParameterStatus(conn, "client_encoding",
val);
}
else
{
/*
* Error: presumably function not available, so
* use PGCLIENTENCODING or SQL_ASCII as the
* fallback.
*/
val = getenv("PGCLIENTENCODING");
if (val && *val)
pqSaveParameterStatus(conn, "client_encoding",
val);
else
pqSaveParameterStatus(conn, "client_encoding",
"SQL_ASCII");
}
PQclear(res);
/* Keep reading until PQgetResult returns NULL */
}
else
{
/* Query finished, so we're done */
conn->setenv_state = SETENV_STATE_IDLE;
return PGRES_POLLING_OK;
}
break;
}
default:
printfPQExpBuffer(&conn->errorMessage,
libpq_gettext("invalid state %c, "
"probably indicative of memory corruption\n"),
conn->setenv_state);
goto error_return;
}
}
/* Unreachable */
error_return:
conn->setenv_state = SETENV_STATE_IDLE;
return PGRES_POLLING_FAILED;
}
/*
* PQendcopy
*
* See fe-exec.c for documentation.
*/
int
pqEndcopy2(PGconn *conn)
{
PGresult *result;
if (conn->asyncStatus != PGASYNC_COPY_IN &&
conn->asyncStatus != PGASYNC_COPY_OUT)
{
printfPQExpBuffer(&conn->errorMessage,
libpq_gettext("no COPY in progress\n"));
return 1;
}
/*
* make sure no data is waiting to be sent, abort if we are non-blocking
* and the flush fails
*/
if (pqFlush(conn) && pqIsnonblocking(conn))
return 1;
/* non blocking connections may have to abort at this point. */
if (pqIsnonblocking(conn) && PQisBusy(conn))
return 1;
/* Return to active duty */
conn->asyncStatus = PGASYNC_BUSY;
resetPQExpBuffer(&conn->errorMessage);
/* Wait for the completion response */
result = PQgetResult(conn);
/* Expecting a successful result */
if (result && result->resultStatus == PGRES_COMMAND_OK)
{
PQclear(result);
return 0;
}
/*
* Trouble. For backwards-compatibility reasons, we issue the error
* message as if it were a notice (would be nice to get rid of this
* silliness, but too many apps probably don't handle errors from
* PQendcopy reasonably). Note that the app can still obtain the error
* status from the PGconn object.
*/
if (conn->errorMessage.len > 0)
{
/* We have to strip the trailing newline ... pain in neck... */
char svLast = conn->errorMessage.data[conn->errorMessage.len - 1];
if (svLast == '\n')
conn->errorMessage.data[conn->errorMessage.len - 1] = '\0';
pqInternalNotice(&conn->noticeHooks, "%s", conn->errorMessage.data);
conn->errorMessage.data[conn->errorMessage.len - 1] = svLast;
}
PQclear(result);
/*
* The worst case is that we've lost sync with the backend entirely due to
* application screwup of the copy in/out protocol. To recover, reset the
* connection (talk about using a sledgehammer...)
*/
pqInternalNotice(&conn->noticeHooks,
"lost synchronization with server, resetting connection");
/*
* Users doing non-blocking connections need to handle the reset
* themselves, they'll need to check the connection status if we return an
* error.
*/
if (pqIsnonblocking(conn))
PQresetStart(conn);
else
PQreset(conn);
return 1;
}
/* ----------
* remoteListen_receive_events
*
* Retrieve all new events that origin from nodes for which we listen on this
* node as provider and add them to the node specific worker message queue.
* ----------
*/
static int
remoteListen_receive_events(SlonNode * node, SlonConn * conn,
struct listat * listat)
{
SlonNode *origin;
SlonDString query;
SlonDString q2;
char *where_or_or;
char seqno_buf[64];
PGresult *res;
int ntuples;
int tupno;
time_t timeout;
time_t now;
dstring_init(&query);
/*
* In the runtime configuration info for the node, we remember the last
* event sequence that we actually have received. If the remote worker
* thread has processed it yet or it isn't important, we have it in the
* message queue at least and don't need to select it again.
*
* So the query we construct contains a qualification (ev_origin =
* <remote_node> and ev_seqno > <last_seqno>) per remote node we're listen
* for here.
*/
monitor_state("remote listener", node->no_id, conn->conn_pid, "receiving events", 0, "n/a");
(void) slon_mkquery(&query,
"select ev_origin, ev_seqno, ev_timestamp, "
" ev_snapshot, "
" \"pg_catalog\".txid_snapshot_xmin(ev_snapshot), "
" \"pg_catalog\".txid_snapshot_xmax(ev_snapshot), "
" ev_type, "
" ev_data1, ev_data2, "
" ev_data3, ev_data4, "
" ev_data5, ev_data6, "
" ev_data7, ev_data8 "
"from %s.sl_event e",
rtcfg_namespace);
rtcfg_lock();
where_or_or = "where";
if (lag_interval)
{
dstring_init(&q2);
(void) slon_mkquery(&q2, "where ev_timestamp < now() - '%s'::interval and (", lag_interval);
where_or_or = dstring_data(&q2);
}
while (listat)
{
if ((origin = rtcfg_findNode(listat->li_origin)) == NULL)
{
rtcfg_unlock();
slon_log(SLON_ERROR,
"remoteListenThread_%d: unknown node %d\n",
node->no_id, listat->li_origin);
dstring_free(&query);
return -1;
}
sprintf(seqno_buf, INT64_FORMAT, origin->last_event);
slon_appendquery(&query,
" %s (e.ev_origin = '%d' and e.ev_seqno > '%s')",
where_or_or, listat->li_origin, seqno_buf);
where_or_or = "or";
listat = listat->next;
}
if (lag_interval)
{
slon_appendquery(&query, ")");
}
/*
* Limit the result set size to: sync_group_maxsize * 2, if it's set 100,
* if sync_group_maxsize isn't set
*/
slon_appendquery(&query, " order by e.ev_origin, e.ev_seqno limit %d",
(sync_group_maxsize > 0) ? sync_group_maxsize * 2 : 100);
rtcfg_unlock();
if (PQsendQuery(conn->dbconn, dstring_data(&query)) == 0)
{
slon_log(SLON_ERROR,
"remoteListenThread_%d: \"%s\" - %s",
node->no_id,
dstring_data(&query), PQerrorMessage(conn->dbconn));
dstring_free(&query);
return -1;
}
(void) time(&timeout);
timeout += remote_listen_timeout;
while (PQisBusy(conn->dbconn) != 0)
{
(void) time(&now);
if (now >= timeout)
{
slon_log(SLON_ERROR,
"remoteListenThread_%d: timeout (%d s) for event selection\n",
node->no_id, remote_listen_timeout);
dstring_free(&query);
return -1;
}
if (PQconsumeInput(conn->dbconn) == 0)
{
slon_log(SLON_ERROR,
"remoteListenThread_%d: \"%s\" - %s",
node->no_id,
dstring_data(&query), PQerrorMessage(conn->dbconn));
dstring_free(&query);
return -1;
}
if (PQisBusy(conn->dbconn) != 0)
sched_wait_time(conn, SCHED_WAIT_SOCK_READ, 10000);
}
res = PQgetResult(conn->dbconn);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
slon_log(SLON_ERROR,
"remoteListenThread_%d: \"%s\" - %s",
node->no_id,
dstring_data(&query), PQresultErrorMessage(res));
PQclear(res);
dstring_free(&query);
return -1;
}
dstring_free(&query);
/*
* Add all events found to the remote worker message queue.
*/
ntuples = PQntuples(res);
/* If we drew in the maximum number of events */
if (ntuples == ((sync_group_maxsize > 0) ? sync_group_maxsize * 2 : 100))
sel_max_events++; /* Add to the count... */
else
sel_max_events = 0; /* reset the count */
for (tupno = 0; tupno < ntuples; tupno++)
{
int ev_origin;
int64 ev_seqno;
ev_origin = (int) strtol(PQgetvalue(res, tupno, 0), NULL, 10);
(void) slon_scanint64(PQgetvalue(res, tupno, 1), &ev_seqno);
slon_log(SLON_DEBUG2, "remoteListenThread_%d: "
"queue event %d,%s %s\n",
node->no_id, ev_origin, PQgetvalue(res, tupno, 1),
PQgetvalue(res, tupno, 6));
remoteWorker_event(node->no_id,
ev_origin, ev_seqno,
PQgetvalue(res, tupno, 2), /* ev_timestamp */
PQgetvalue(res, tupno, 3), /* ev_snapshot */
PQgetvalue(res, tupno, 4), /* mintxid */
PQgetvalue(res, tupno, 5), /* maxtxid */
PQgetvalue(res, tupno, 6), /* ev_type */
(PQgetisnull(res, tupno, 7)) ? NULL : PQgetvalue(res, tupno, 7),
(PQgetisnull(res, tupno, 8)) ? NULL : PQgetvalue(res, tupno, 8),
(PQgetisnull(res, tupno, 9)) ? NULL : PQgetvalue(res, tupno, 9),
(PQgetisnull(res, tupno, 10)) ? NULL : PQgetvalue(res, tupno, 10),
(PQgetisnull(res, tupno, 11)) ? NULL : PQgetvalue(res, tupno, 11),
(PQgetisnull(res, tupno, 12)) ? NULL : PQgetvalue(res, tupno, 12),
(PQgetisnull(res, tupno, 13)) ? NULL : PQgetvalue(res, tupno, 13),
(PQgetisnull(res, tupno, 14)) ? NULL : PQgetvalue(res, tupno, 14));
}
if (ntuples > 0)
{
if ((sel_max_events > 2) && (sync_group_maxsize > 100))
{
slon_log(SLON_INFO, "remoteListenThread_%d: drew maximum # of events for %d iterations\n",
node->no_id, sel_max_events);
sched_msleep(node, 10000 + (1000 * sel_max_events));
}
else
{
poll_sleep = 0;
}
}
else
{
poll_sleep = poll_sleep * 2 + sync_interval;
if (poll_sleep > sync_interval_timeout)
{
poll_sleep = sync_interval_timeout;
}
}
PQclear(res);
monitor_state("remote listener", node->no_id, conn->conn_pid, "thread main loop", 0, "n/a");
return 0;
}
int pgQueryThread::execute()
{
rowsInserted = -1L;
if (!conn->conn)
return(raiseEvent(0));
wxCharBuffer queryBuf = query.mb_str(*conn->conv);
if (!queryBuf && !query.IsEmpty())
{
conn->SetLastResultError(NULL, _("The query could not be converted to the required encoding."));
return(raiseEvent(0));
}
if (!PQsendQuery(conn->conn, queryBuf))
{
conn->SetLastResultError(NULL);
conn->IsAlive();
return(raiseEvent(0));
}
int resultsRetrieved = 0;
PGresult *lastResult = 0;
while (true)
{
if (TestDestroy())
{
if (rc != -3)
{
if (!PQrequestCancel(conn->conn)) // could not abort; abort failed.
return(raiseEvent(-1));
rc = -3;
}
}
if (!PQconsumeInput(conn->conn))
return(raiseEvent(0));
if (PQisBusy(conn->conn))
{
Yield();
this->Sleep(10);
continue;
}
// If resultToRetrieve is given, the nth result will be returned,
// otherwise the last result set will be returned.
// all others are discarded
PGresult *res = PQgetResult(conn->conn);
if (!res)
break;
resultsRetrieved++;
if (resultsRetrieved == resultToRetrieve)
{
result = res;
insertedOid = PQoidValue(res);
if (insertedOid && insertedOid != (OID) - 1)
appendMessage(wxString::Format(_("Query inserted one row with OID %d.\n"), insertedOid));
else
appendMessage(wxString::Format(wxPLURAL("Query result with %d row will be returned.\n", "Query result with %d rows will be returned.\n",
PQntuples(result))));
continue;
}
if (lastResult)
{
if (PQntuples(lastResult))
appendMessage(wxString::Format(wxPLURAL("Query result with %d row discarded.\n", "Query result with %d rows discarded.\n",
PQntuples(lastResult))));
PQclear(lastResult);
}
lastResult = res;
}
if (!result)
result = lastResult;
conn->SetLastResultError(result);
appendMessage(wxT("\n"));
rc = PQresultStatus(result);
insertedOid = PQoidValue(result);
if (insertedOid == (OID) - 1)
insertedOid = 0;
if (rc == PGRES_TUPLES_OK)
{
dataSet = new pgSet(result, conn, *conn->conv, conn->needColQuoting);
dataSet->MoveFirst();
}
else if (rc == PGRES_COMMAND_OK)
{
char *s = PQcmdTuples(result);
if (*s)
rowsInserted = atol(s);
}
else if (rc == PGRES_FATAL_ERROR)
{
appendMessage(conn->GetLastError() + wxT("\n"));
}
return(raiseEvent(1));
}
/*
* GetIdleSlot
* Return a connection slot that is ready to execute a command.
*
* We return the first slot we find that is marked isFree, if one is;
* otherwise, we loop on select() until one socket becomes available. When
* this happens, we read the whole set and mark as free all sockets that become
* available.
*
* Process the slot list, if any free slot is available then return the slotid
* else perform the select on all the socket's and wait until at least one slot
* becomes available.
*
* If an error occurs, NULL is returned.
*/
static ParallelSlot *
GetIdleSlot(ParallelSlot slots[], int numslots, const char *dbname,
const char *progname)
{
int i;
int firstFree = -1;
fd_set slotset;
pgsocket maxFd;
for (i = 0; i < numslots; i++)
if ((slots + i)->isFree)
return slots + i;
FD_ZERO(&slotset);
maxFd = slots->sock;
for (i = 0; i < numslots; i++)
{
FD_SET((slots + i)->sock, &slotset);
if ((slots + i)->sock > maxFd)
maxFd = (slots + i)->sock;
}
/*
* No free slot found, so wait until one of the connections has finished
* its task and return the available slot.
*/
for (firstFree = -1; firstFree < 0;)
{
bool aborting;
SetCancelConn(slots->connection);
i = select_loop(maxFd, &slotset, &aborting);
ResetCancelConn();
if (aborting)
{
/*
* We set the cancel-receiving connection to the one in the zeroth
* slot above, so fetch the error from there.
*/
GetQueryResult(slots->connection, dbname, progname);
return NULL;
}
Assert(i != 0);
for (i = 0; i < numslots; i++)
{
if (!FD_ISSET((slots + i)->sock, &slotset))
continue;
PQconsumeInput((slots + i)->connection);
if (PQisBusy((slots + i)->connection))
continue;
(slots + i)->isFree = true;
if (!GetQueryResult((slots + i)->connection, dbname, progname))
return NULL;
if (firstFree < 0)
firstFree = i;
}
}
return slots + firstFree;
}
/*
* Our caller already sent the query associated with this step. Wait for it
* to either complete or (if given the STEP_NONBLOCK flag) to block while
* waiting for a lock. We assume that any lock wait will persist until we
* have executed additional steps in the permutation.
*
* When calling this function on behalf of a given step for a second or later
* time, pass the STEP_RETRY flag. This only affects the messages printed.
*
* If the connection returns an error, the message is saved in step->errormsg.
* Caller should call report_error_message shortly after this, to have it
* printed and cleared.
*
* If the STEP_NONBLOCK flag was specified and the query is waiting to acquire
* a lock, returns true. Otherwise, returns false.
*/
static bool
try_complete_step(Step * step, int flags)
{
PGconn *conn = conns[1 + step->session];
fd_set read_set;
struct timeval timeout;
int sock = PQsocket(conn);
int ret;
PGresult *res;
FD_ZERO(&read_set);
while ((flags & STEP_NONBLOCK) && PQisBusy(conn))
{
FD_SET(sock, &read_set);
timeout.tv_sec = 0;
timeout.tv_usec = 10000; /* Check for lock waits every 10ms. */
ret = select(sock + 1, &read_set, NULL, NULL, &timeout);
if (ret < 0) /* error in select() */
{
if (errno == EINTR)
continue;
fprintf(stderr, "select failed: %s\n", strerror(errno));
exit_nicely();
}
else if (ret == 0) /* select() timeout: check for lock wait */
{
int ntuples;
res = PQexecPrepared(conns[0], PREP_WAITING, 1,
&backend_pids[step->session + 1],
NULL, NULL, 0);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
fprintf(stderr, "lock wait query failed: %s",
PQerrorMessage(conn));
exit_nicely();
}
ntuples = PQntuples(res);
PQclear(res);
if (ntuples >= 1) /* waiting to acquire a lock */
{
if (!(flags & STEP_RETRY))
printf("step %s: %s <waiting ...>\n",
step->name, step->sql);
return true;
}
/* else, not waiting: give it more time */
}
else if (!PQconsumeInput(conn)) /* select(): data available */
{
fprintf(stderr, "PQconsumeInput failed: %s\n",
PQerrorMessage(conn));
exit_nicely();
}
}
if (flags & STEP_RETRY)
printf("step %s: <... completed>\n", step->name);
else
printf("step %s: %s\n", step->name, step->sql);
while ((res = PQgetResult(conn)))
{
switch (PQresultStatus(res))
{
case PGRES_COMMAND_OK:
break;
case PGRES_TUPLES_OK:
printResultSet(res);
break;
case PGRES_FATAL_ERROR:
if (step->errormsg != NULL)
{
printf("WARNING: this step had a leftover error message\n");
printf("%s\n", step->errormsg);
}
/*
* Detail may contain XID values, so we want to just show
* primary. Beware however that libpq-generated error results
* may not contain subfields, only an old-style message.
*/
{
const char *sev = PQresultErrorField(res,
PG_DIAG_SEVERITY);
const char *msg = PQresultErrorField(res,
PG_DIAG_MESSAGE_PRIMARY);
if (sev && msg)
{
step->errormsg = malloc(5 + strlen(sev) + strlen(msg));
sprintf(step->errormsg, "%s: %s", sev, msg);
}
else
step->errormsg = strdup(PQresultErrorMessage(res));
}
break;
default:
printf("unexpected result status: %s\n",
PQresStatus(PQresultStatus(res)));
}
PQclear(res);
}
return false;
}
/*
* Insert monitor info, this is basically the time and xlog replayed,
* applied on standby and current xlog location in primary.
* Also do the math to see how far are we in bytes for being uptodate
*/
static void
MonitorExecute(void)
{
PGresult *res;
char monitor_standby_timestamp[MAXLEN];
char last_wal_primary_location[MAXLEN];
char last_wal_standby_received[MAXLEN];
char last_wal_standby_applied[MAXLEN];
unsigned long long int lsn_primary;
unsigned long long int lsn_standby_received;
unsigned long long int lsn_standby_applied;
int connection_retries;
/*
* Check if the master is still available, if after 5 minutes of retries
* we cannot reconnect, try to get a new master.
*/
for (connection_retries = 0; connection_retries < 15; connection_retries++)
{
if (PQstatus(primaryConn) != CONNECTION_OK)
{
log_warning(_("Connection to master has been lost, trying to recover...\n"));
/* wait 20 seconds between retries */
sleep(20);
PQreset(primaryConn);
}
else
{
if (connection_retries > 0)
{
log_notice(_("Connection to master has been restored, continue monitoring.\n"));
}
break;
}
}
if (PQstatus(primaryConn) != CONNECTION_OK)
{
log_err(_("We couldn't reconnect to master. Now checking if another node has been promoted.\n"));
for (connection_retries = 0; connection_retries < 6; connection_retries++)
{
primaryConn = getMasterConnection(myLocalConn, local_options.node,
local_options.cluster_name, &primary_options.node,NULL);
if (PQstatus(primaryConn) == CONNECTION_OK)
{
/* Connected, we can continue the process so break the loop */
log_err(_("Connected to node %d, continue monitoring.\n"), primary_options.node);
break;
}
else
{
log_err(_("We haven't found a new master, waiting before retry...\n"));
/* wait 5 minutes before retries, after 6 failures (30 minutes) we stop trying */
sleep(300);
}
}
}
if (PQstatus(primaryConn) != CONNECTION_OK)
{
log_err(_("We couldn't reconnect for long enough, exiting...\n"));
exit(ERR_DB_CON);
}
/* Check if we still are a standby, we could have been promoted */
if (!is_standby(myLocalConn))
{
log_err(_("It seems like we have been promoted, so exit from monitoring...\n"));
CloseConnections();
exit(ERR_PROMOTED);
}
/*
* first check if there is a command being executed,
* and if that is the case, cancel the query so i can
* insert the current record
*/
if (PQisBusy(primaryConn) == 1)
CancelQuery();
/* Get local xlog info */
sqlquery_snprintf(
sqlquery,
"SELECT CURRENT_TIMESTAMP, pg_last_xlog_receive_location(), "
"pg_last_xlog_replay_location()");
res = PQexec(myLocalConn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err("PQexec failed: %s\n", PQerrorMessage(myLocalConn));
PQclear(res);
/* if there is any error just let it be and retry in next loop */
return;
}
strncpy(monitor_standby_timestamp, PQgetvalue(res, 0, 0), MAXLEN);
strncpy(last_wal_standby_received , PQgetvalue(res, 0, 1), MAXLEN);
strncpy(last_wal_standby_applied , PQgetvalue(res, 0, 2), MAXLEN);
PQclear(res);
/* Get primary xlog info */
sqlquery_snprintf(sqlquery, "SELECT pg_current_xlog_location() ");
res = PQexec(primaryConn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err("PQexec failed: %s\n", PQerrorMessage(primaryConn));
PQclear(res);
return;
}
strncpy(last_wal_primary_location, PQgetvalue(res, 0, 0), MAXLEN);
PQclear(res);
/* Calculate the lag */
lsn_primary = walLocationToBytes(last_wal_primary_location);
lsn_standby_received = walLocationToBytes(last_wal_standby_received);
lsn_standby_applied = walLocationToBytes(last_wal_standby_applied);
if (only_one_entry && only_one_entry_desired)
{
sqlquery_snprintf(sqlquery,
"UPDATE %s.repl_monitor "
"VALUES(%d, %d, '%s'::timestamp with time zone, "
" '%s', '%s', "
" %lld, %lld)"
"WHERE primary_node=%d AND secondary_node=%d", repmgr_schema,
primary_options.node, local_options.node, monitor_standby_timestamp,
last_wal_primary_location,
last_wal_standby_received,
(lsn_primary - lsn_standby_received),
(lsn_standby_received - lsn_standby_applied));
res = PQexec(primaryConn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err("PQexec failed: %s\n", PQerrorMessage(conn));
PQclear(res);
CloseConnections();
exit(ERR_DB_QUERY);
}
if (PQntuples(res) != 1)
{
only_one_entry = false;
}
PQclear(res);
}
else
{
/*
* Build and send insert
*/
sqlquery_snprintf(sqlquery,
"INSERT INTO %s.repl_monitor "
"VALUES(%d, %d, '%s'::timestamp with time zone, "
" '%s', '%s', "
" %lld, %lld)", repmgr_schema,
primary_options.node, local_options.node, monitor_standby_timestamp,
last_wal_primary_location,
last_wal_standby_received,
(lsn_primary - lsn_standby_received),
(lsn_standby_received - lsn_standby_applied));
res = PQexec(primaryConn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err("PQexec failed: %s\n", PQerrorMessage(conn));
PQclear(res);
CloseConnections();
exit(ERR_DB_QUERY);
}
PQclear(res);
if (only_one_entry_desired)
{
/*
* Build the SQL to execute on primary
*/
sqlquery_snprintf(sqlquery,
"DELETE FROM %s.repl_monitor "
"WHERE primary_node=%d AND standby_node=%d AND last_monitor_time < '%s'::timestamp with time zone",
repmgr_schema, primary_options.node, local_options.node, monitor_standby_timestamp);
res = PQexec(primaryConn, sqlquery);
if (PQresultStatus(res) != PGRES_TUPLES_OK)
{
log_err("PQexec failed: %s\n", PQerrorMessage(conn));
PQclear(res);
CloseConnections();
exit(ERR_DB_QUERY);
}
PQclear(res);
only_one_entry = true;
}
}
}