/*
* Execute the CREATE BARRIER command. Write a BARRIER WAL record and flush the
* WAL buffers to disk before returning to the caller. Writing the WAL record
* does not guarantee successful completion of the barrier command.
*/
void
ProcessCreateBarrierExecute(const char *id)
{
StringInfoData buf;
if (!IsConnFromCoord())
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("The CREATE BARRIER EXECUTE message is expected to "
"arrive from a Coordinator")));
{
XLogRecData rdata[1];
XLogRecPtr recptr;
rdata[0].data = (char *) id;
rdata[0].len = strlen(id) + 1;
rdata[0].buffer = InvalidBuffer;
rdata[0].next = NULL;
recptr = XLogInsert(RM_BARRIER_ID, XLOG_BARRIER_CREATE, rdata);
XLogFlush(recptr);
}
pq_beginmessage(&buf, 'b');
pq_sendstring(&buf, id);
pq_endmessage(&buf);
pq_flush();
}
/* ----------------
* ReadyForQuery - tell dest that we are ready for a new query
*
* The ReadyForQuery message is sent in protocol versions 2.0 and up
* so that the FE can tell when we are done processing a query string.
* In versions 3.0 and up, it also carries a transaction state indicator.
*
* Note that by flushing the stdio buffer here, we can avoid doing it
* most other places and thus reduce the number of separate packets sent.
* ----------------
*/
void
ReadyForQuery(CommandDest dest)
{
switch (dest)
{
case DestRemote:
case DestRemoteExecute:
if (PG_PROTOCOL_MAJOR(FrontendProtocol) >= 3)
{
StringInfoData buf;
pq_beginmessage(&buf, 'Z');
pq_sendbyte(&buf, TransactionBlockStatusCode());
pq_endmessage(&buf);
}
else if (PG_PROTOCOL_MAJOR(FrontendProtocol) >= 2)
pq_putemptymessage('Z');
/* Flush output at end of cycle in any case. */
pq_flush();
break;
case DestNone:
case DestDebug:
case DestSPI:
case DestTuplestore:
case DestIntoRel:
case DestCopyOut:
break;
}
}
static Source *
CreateRemoteSource(const char *path, TupleDesc desc)
{
RemoteSource *self = (RemoteSource *) palloc0(sizeof(RemoteSource));
self->base.close = (SourceCloseProc) RemoteSourceClose;
if (PG_PROTOCOL_MAJOR(FrontendProtocol) >= 3)
{
/* new way */
StringInfoData buf;
int16 format;
int nattrs;
int i;
self->base.read = (SourceReadProc) RemoteSourceRead;
/* count valid fields */
for (nattrs = 0, i = 0; i < desc->natts; i++)
{
if (desc->attrs[i]->attisdropped)
continue;
nattrs++;
}
format = (IsBinaryCopy() ? 1 : 0);
pq_beginmessage(&buf, 'G');
pq_sendbyte(&buf, format); /* overall format */
pq_sendint(&buf, nattrs, 2);
for (i = 0; i < nattrs; i++)
pq_sendint(&buf, format, 2); /* per-column formats */
pq_endmessage(&buf);
self->buffer = makeStringInfo();
}
else if (PG_PROTOCOL_MAJOR(FrontendProtocol) >= 2)
{
self->base.read = (SourceReadProc) RemoteSourceReadOld;
/* old way */
if (IsBinaryCopy())
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("COPY BINARY is not supported to stdout or from stdin")));
pq_putemptymessage('G');
}
else
{
self->base.read = (SourceReadProc) RemoteSourceReadOld;
/* very old way */
if (IsBinaryCopy())
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("COPY BINARY is not supported to stdout or from stdin")));
pq_putemptymessage('D');
}
/* We *must* flush here to ensure FE knows it can send. */
pq_flush();
return (Source *) self;
}
static void
putEndLocationReply(XLogRecPtr *endLocation)
{
StringInfoData buf;
pq_beginmessage(&buf, 's');
pq_sendint(&buf, endLocation->xlogid, 4);
pq_sendint(&buf, endLocation->xrecoff, 4);
pq_endmessage(&buf);
pq_flush();
}
/*
* START_REPLICATION
*/
static void
StartReplication(StartReplicationCmd *cmd)
{
StringInfoData buf;
/*
* Let postmaster know that we're streaming. Once we've declared us as a
* WAL sender process, postmaster will let us outlive the bgwriter and
* kill us last in the shutdown sequence, so we get a chance to stream all
* remaining WAL at shutdown, including the shutdown checkpoint. Note that
* there's no going back, and we mustn't write any WAL records after this.
*/
MarkPostmasterChildWalSender();
SendPostmasterSignal(PMSIGNAL_ADVANCE_STATE_MACHINE);
/*
* Check that we're logging enough information in the WAL for
* log-shipping.
*
* NOTE: This only checks the current value of wal_level. Even if the
* current setting is not 'minimal', there can be old WAL in the pg_xlog
* directory that was created with 'minimal'. So this is not bulletproof,
* the purpose is just to give a user-friendly error message that hints
* how to configure the system correctly.
*/
if (wal_level == WAL_LEVEL_MINIMAL)
ereport(FATAL,
(errcode(ERRCODE_CANNOT_CONNECT_NOW),
errmsg("standby connections not allowed because wal_level=minimal")));
/*
* When we first start replication the standby will be behind the primary.
* For some applications, for example, synchronous replication, it is
* important to have a clear state for this initial catchup mode, so we
* can trigger actions when we change streaming state later. We may stay
* in this state for a long time, which is exactly why we want to be able
* to monitor whether or not we are still here.
*/
WalSndSetState(WALSNDSTATE_CATCHUP);
/* Send a CopyBothResponse message, and start streaming */
pq_beginmessage(&buf, 'W');
pq_sendbyte(&buf, 0);
pq_sendint(&buf, 0, 2);
pq_endmessage(&buf);
pq_flush();
/*
* Initialize position to the received one, then the xlog records begin to
* be shipped from that position
*/
sentPtr = cmd->startpoint;
}
/*
* START_REPLICATION
*/
static void
StartReplication(StartReplicationCmd * cmd)
{
StringInfoData buf;
/*
* Let postmaster know that we're streaming. Once we've declared us as
* a WAL sender process, postmaster will let us outlive the bgwriter and
* kill us last in the shutdown sequence, so we get a chance to stream
* all remaining WAL at shutdown, including the shutdown checkpoint.
* Note that there's no going back, and we mustn't write any WAL records
* after this.
*/
MarkPostmasterChildWalSender();
/*
* Check that we're logging enough information in the WAL for
* log-shipping.
*
* NOTE: This only checks the current value of wal_level. Even if the
* current setting is not 'minimal', there can be old WAL in the pg_xlog
* directory that was created with 'minimal'. So this is not bulletproof,
* the purpose is just to give a user-friendly error message that hints
* how to configure the system correctly.
*/
if (wal_level == WAL_LEVEL_MINIMAL)
ereport(FATAL,
(errcode(ERRCODE_CANNOT_CONNECT_NOW),
errmsg("standby connections not allowed because wal_level=minimal")));
/* Send a CopyBothResponse message, and start streaming */
pq_beginmessage(&buf, 'W');
pq_sendbyte(&buf, 0);
pq_sendint(&buf, 0, 2);
pq_endmessage(&buf);
pq_flush();
/*
* Initialize position to the received one, then the xlog records begin to
* be shipped from that position
*/
sentPtr = cmd->startpoint;
}
/*
* Send an authentication request packet to the frontend.
*/
static void
sendAuthRequest(Port *port, AuthRequest areq)
{
StringInfoData buf;
pq_beginmessage(&buf, 'R');
pq_sendint(&buf, (int32) areq, sizeof(int32));
/* Add the salt for encrypted passwords. */
if (areq == AUTH_REQ_MD5)
pq_sendbytes(&buf, port->md5Salt, 4);
else if (areq == AUTH_REQ_CRYPT)
pq_sendbytes(&buf, port->cryptSalt, 2);
#if defined(ENABLE_GSS) || defined(ENABLE_SSPI)
/*
* Add the authentication data for the next step of the GSSAPI or SSPI
* negotiation.
*/
else if (areq == AUTH_REQ_GSS_CONT)
{
if (port->gss->outbuf.length > 0)
{
elog(DEBUG4, "sending GSS token of length %u",
(unsigned int) port->gss->outbuf.length);
pq_sendbytes(&buf, port->gss->outbuf.value, port->gss->outbuf.length);
}
}
#endif
pq_endmessage(&buf);
/*
* Flush message so client will see it, except for AUTH_REQ_OK, which need
* not be sent until we are ready for queries.
*/
if (areq != AUTH_REQ_OK)
pq_flush();
}
void
ProcessGTMEndBackup(Port *myport, StringInfo message)
{
int ii;
GTM_ThreadInfo *my_threadinfo;
StringInfoData buf;
pq_getmsgend(message);
my_threadinfo = GetMyThreadInfo;
for (ii = 0; ii < GTMThreads->gt_array_size; ii++)
{
if (GTMThreads->gt_threads[ii] && GTMThreads->gt_threads[ii] != my_threadinfo)
GTM_RWLockRelease(>MThreads->gt_threads[ii]->thr_lock);
}
my_threadinfo->thr_status = GTM_THREAD_RUNNING;
pq_beginmessage(&buf, 'S');
pq_sendint(&buf, END_BACKUP_RESULT, 4);
pq_endmessage(myport, &buf);
pq_flush(myport);
}
void
ProcessGTMBeginBackup(Port *myport, StringInfo message)
{
int ii;
GTM_ThreadInfo *my_threadinfo;
StringInfoData buf;
pq_getmsgend(message);
my_threadinfo = GetMyThreadInfo;
for (ii = 0; ii < GTMThreads->gt_array_size; ii++)
{
if (GTMThreads->gt_threads[ii] && GTMThreads->gt_threads[ii] != my_threadinfo)
GTM_RWLockAcquire(>MThreads->gt_threads[ii]->thr_lock, GTM_LOCKMODE_WRITE);
}
my_threadinfo->thr_status = GTM_THREAD_BACKUP;
pq_beginmessage(&buf, 'S');
pq_sendint(&buf, BEGIN_BACKUP_RESULT, 4);
pq_endmessage(myport, &buf);
pq_flush(myport);
}
static void
send_buffer()
{
if (buffer_len > 0)
{
StringInfoData msgbuf;
char *cursor = buffer;
while (--buffer_len > 0)
{
if (*cursor == '\0')
*cursor = '\n';
cursor++;
}
if (*cursor != '\0')
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("internal error"),
errdetail("Wrong message format detected")));
pq_beginmessage(&msgbuf, 'N');
if (PG_PROTOCOL_MAJOR(FrontendProtocol) >= 3)
{
pq_sendbyte(&msgbuf, PG_DIAG_MESSAGE_PRIMARY);
pq_sendstring(&msgbuf, buffer);
pq_sendbyte(&msgbuf, '\0');
}
else
{
*cursor++ = '\n';
*cursor = '\0';
pq_sendstring(&msgbuf, buffer);
}
pq_endmessage(&msgbuf);
pq_flush();
}
}
/*
* Mark the completion of an on-going barrier. We must have remembered the
* barrier ID when we received the CREATE BARRIER PREPARE command
*/
void
ProcessCreateBarrierEnd(const char *id)
{
StringInfoData buf;
if (!IS_PGXC_COORDINATOR || !IsConnFromCoord())
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("The CREATE BARRIER END message is expected to "
"arrive at a Coordinator from another Coordinator")));
LWLockRelease(BarrierLock);
pq_beginmessage(&buf, 'b');
pq_sendstring(&buf, id);
pq_endmessage(&buf);
pq_flush();
/*
* TODO Stop the timer
*/
}
/*
* Prepare ourselves for an incoming BARRIER. We must disable all new 2PC
* commits and let the ongoing commits to finish. We then remember the
* barrier id (so that it can be matched with the final END message) and
* tell the driving Coordinator to proceed with the next step.
*
* A simple way to implement this is to grab a lock in an exclusive mode
* while all other backend starting a 2PC will grab the lock in shared
* mode. So as long as we hold the exclusive lock, no other backend start a
* new 2PC and there can not be any 2PC in-progress. This technique would
* rely on assumption that an exclusive lock requester is not starved by
* share lock requesters.
*
* Note: To ensure that the 2PC are not blocked for a long time, we should
* set a timeout. The lock should be release after the timeout and the
* barrier should be canceled.
*/
void
ProcessCreateBarrierPrepare(const char *id)
{
StringInfoData buf;
if (!IS_PGXC_COORDINATOR || !IsConnFromCoord())
ereport(ERROR,
(errcode(ERRCODE_INTERNAL_ERROR),
errmsg("The CREATE BARRIER PREPARE message is expected to "
"arrive at a Coordinator from another Coordinator")));
LWLockAcquire(BarrierLock, LW_EXCLUSIVE);
pq_beginmessage(&buf, 'b');
pq_sendstring(&buf, id);
pq_endmessage(&buf);
pq_flush();
/*
* TODO Start a timer to terminate the pending barrier after a specified
* timeout
*/
}
/*
* Send an authentication request packet to the frontend.
*/
static void
sendAuthRequest(Port *port, AuthRequest areq)
{
StringInfoData buf;
pq_beginmessage(&buf, 'R');
pq_sendint(&buf, (int32) areq, sizeof(int32));
/* Add the salt for encrypted passwords. */
if (areq == AUTH_REQ_MD5)
pq_sendbytes(&buf, port->md5Salt, 4);
else if (areq == AUTH_REQ_CRYPT)
pq_sendbytes(&buf, port->cryptSalt, 2);
pq_endmessage(&buf);
/*
* Flush message so client will see it, except for AUTH_REQ_OK, which
* need not be sent until we are ready for queries.
*/
if (areq != AUTH_REQ_OK)
pq_flush();
}
/*
* Execute commands from walreceiver, until we enter streaming mode.
*/
static void
WalSndHandshake(void)
{
StringInfoData input_message;
bool replication_started = false;
initStringInfo(&input_message);
while (!replication_started)
{
int firstchar;
/* Wait for a command to arrive */
firstchar = pq_getbyte();
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (!PostmasterIsAlive(true))
exit(1);
/*
* Check for any other interesting events that happened while we
* slept.
*/
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
}
if (firstchar != EOF)
{
/*
* Read the message contents. This is expected to be done without
* blocking because we've been able to get message type code.
*/
if (pq_getmessage(&input_message, 0))
firstchar = EOF; /* suitable message already logged */
}
/* Handle the very limited subset of commands expected in this phase */
switch (firstchar)
{
case 'Q': /* Query message */
{
const char *query_string;
XLogRecPtr recptr;
query_string = pq_getmsgstring(&input_message);
pq_getmsgend(&input_message);
if (strcmp(query_string, "IDENTIFY_SYSTEM") == 0)
{
StringInfoData buf;
char sysid[32];
char tli[11];
/*
* Reply with a result set with one row, two columns.
* First col is system ID, and second is timeline ID
*/
snprintf(sysid, sizeof(sysid), UINT64_FORMAT,
GetSystemIdentifier());
snprintf(tli, sizeof(tli), "%u", ThisTimeLineID);
/* Send a RowDescription message */
pq_beginmessage(&buf, 'T');
pq_sendint(&buf, 2, 2); /* 2 fields */
/* first field */
pq_sendstring(&buf, "systemid"); /* col name */
pq_sendint(&buf, 0, 4); /* table oid */
pq_sendint(&buf, 0, 2); /* attnum */
pq_sendint(&buf, TEXTOID, 4); /* type oid */
pq_sendint(&buf, -1, 2); /* typlen */
pq_sendint(&buf, 0, 4); /* typmod */
pq_sendint(&buf, 0, 2); /* format code */
/* second field */
pq_sendstring(&buf, "timeline"); /* col name */
pq_sendint(&buf, 0, 4); /* table oid */
pq_sendint(&buf, 0, 2); /* attnum */
pq_sendint(&buf, INT4OID, 4); /* type oid */
pq_sendint(&buf, 4, 2); /* typlen */
pq_sendint(&buf, 0, 4); /* typmod */
pq_sendint(&buf, 0, 2); /* format code */
pq_endmessage(&buf);
/* Send a DataRow message */
pq_beginmessage(&buf, 'D');
pq_sendint(&buf, 2, 2); /* # of columns */
pq_sendint(&buf, strlen(sysid), 4); /* col1 len */
pq_sendbytes(&buf, (char *) &sysid, strlen(sysid));
pq_sendint(&buf, strlen(tli), 4); /* col2 len */
pq_sendbytes(&buf, (char *) tli, strlen(tli));
pq_endmessage(&buf);
/* Send CommandComplete and ReadyForQuery messages */
EndCommand("SELECT", DestRemote);
ReadyForQuery(DestRemote);
/* ReadyForQuery did pq_flush for us */
}
else if (sscanf(query_string, "START_REPLICATION %X/%X",
&recptr.xlogid, &recptr.xrecoff) == 2)
{
StringInfoData buf;
/*
* Check that we're logging enough information in the
* WAL for log-shipping.
*
* NOTE: This only checks the current value of
* wal_level. Even if the current setting is not
* 'minimal', there can be old WAL in the pg_xlog
* directory that was created with 'minimal'. So this
* is not bulletproof, the purpose is just to give a
* user-friendly error message that hints how to
* configure the system correctly.
*/
if (wal_level == WAL_LEVEL_MINIMAL)
ereport(FATAL,
(errcode(ERRCODE_CANNOT_CONNECT_NOW),
errmsg("standby connections not allowed because wal_level=minimal")));
/* Send a CopyOutResponse message, and start streaming */
pq_beginmessage(&buf, 'H');
pq_sendbyte(&buf, 0);
pq_sendint(&buf, 0, 2);
pq_endmessage(&buf);
pq_flush();
/*
* Initialize position to the received one, then the
* xlog records begin to be shipped from that position
*/
sentPtr = recptr;
/* break out of the loop */
replication_started = true;
}
else
{
ereport(FATAL,
(errcode(ERRCODE_PROTOCOL_VIOLATION),
errmsg("invalid standby query string: %s", query_string)));
}
break;
}
case 'X':
/* standby is closing the connection */
proc_exit(0);
case EOF:
/* standby disconnected unexpectedly */
ereport(COMMERROR,
(errcode(ERRCODE_PROTOCOL_VIOLATION),
errmsg("unexpected EOF on standby connection")));
proc_exit(0);
default:
ereport(FATAL,
(errcode(ERRCODE_PROTOCOL_VIOLATION),
errmsg("invalid standby handshake message type %d", firstchar)));
}
}
}
/* Main loop of walsender process */
static int
WalSndLoop(void)
{
char *output_message;
bool caughtup = false;
/*
* Allocate buffer that will be used for each output message. We do this
* just once to reduce palloc overhead. The buffer must be made large
* enough for maximum-sized messages.
*/
output_message = palloc(1 + sizeof(WalDataMessageHeader) + MAX_SEND_SIZE);
/* Loop forever, unless we get an error */
for (;;)
{
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (!PostmasterIsAlive(true))
exit(1);
/* Process any requests or signals received recently */
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
}
/*
* When SIGUSR2 arrives, we send all outstanding logs up to the
* shutdown checkpoint record (i.e., the latest record) and exit.
*/
if (ready_to_stop)
{
if (!XLogSend(output_message, &caughtup))
break;
if (caughtup)
shutdown_requested = true;
}
/* Normal exit from the walsender is here */
if (shutdown_requested)
{
/* Inform the standby that XLOG streaming was done */
pq_puttextmessage('C', "COPY 0");
pq_flush();
proc_exit(0);
}
/*
* If we had sent all accumulated WAL in last round, nap for the
* configured time before retrying.
*/
if (caughtup)
{
/*
* Even if we wrote all the WAL that was available when we started
* sending, more might have arrived while we were sending this
* batch. We had the latch set while sending, so we have not
* received any signals from that time. Let's arm the latch
* again, and after that check that we're still up-to-date.
*/
ResetLatch(&MyWalSnd->latch);
if (!XLogSend(output_message, &caughtup))
break;
if (caughtup && !got_SIGHUP && !ready_to_stop && !shutdown_requested)
{
/*
* XXX: We don't really need the periodic wakeups anymore,
* WaitLatchOrSocket should reliably wake up as soon as
* something interesting happens.
*/
/* Sleep */
WaitLatchOrSocket(&MyWalSnd->latch, MyProcPort->sock,
WalSndDelay * 1000L);
}
/* Check if the connection was closed */
CheckClosedConnection();
}
else
{
/* Attempt to send the log once every loop */
if (!XLogSend(output_message, &caughtup))
break;
}
}
/*
* Get here on send failure. Clean up and exit.
*
* Reset whereToSendOutput to prevent ereport from attempting to send any
* more messages to the standby.
*/
if (whereToSendOutput == DestRemote)
whereToSendOutput = DestNone;
proc_exit(0);
return 1; /* keep the compiler quiet */
}
/* Main loop of walsender process */
static int
WalSndLoop(void)
{
char *output_message;
bool caughtup = false;
/*
* Allocate buffer that will be used for each output message. We do this
* just once to reduce palloc overhead. The buffer must be made large
* enough for maximum-sized messages.
*/
output_message = palloc(1 + sizeof(WalDataMessageHeader) + MAX_SEND_SIZE);
/*
* Allocate buffer that will be used for processing reply messages. As
* above, do this just once to reduce palloc overhead.
*/
initStringInfo(&reply_message);
/* Initialize the last reply timestamp */
last_reply_timestamp = GetCurrentTimestamp();
/* Loop forever, unless we get an error */
for (;;)
{
/* Clear any already-pending wakeups */
ResetLatch(&MyWalSnd->latch);
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (!PostmasterIsAlive())
exit(1);
/* Process any requests or signals received recently */
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
SyncRepInitConfig();
}
/* Normal exit from the walsender is here */
if (walsender_shutdown_requested)
{
/* Inform the standby that XLOG streaming is done */
pq_puttextmessage('C', "COPY 0");
pq_flush();
proc_exit(0);
}
/* Check for input from the client */
ProcessRepliesIfAny();
/*
* If we don't have any pending data in the output buffer, try to send
* some more. If there is some, we don't bother to call XLogSend
* again until we've flushed it ... but we'd better assume we are not
* caught up.
*/
if (!pq_is_send_pending())
XLogSend(output_message, &caughtup);
else
caughtup = false;
/* Try to flush pending output to the client */
if (pq_flush_if_writable() != 0)
break;
/* If nothing remains to be sent right now ... */
if (caughtup && !pq_is_send_pending())
{
/*
* If we're in catchup state, move to streaming. This is an
* important state change for users to know about, since before
* this point data loss might occur if the primary dies and we
* need to failover to the standby. The state change is also
* important for synchronous replication, since commits that
* started to wait at that point might wait for some time.
*/
if (MyWalSnd->state == WALSNDSTATE_CATCHUP)
{
ereport(DEBUG1,
(errmsg("standby \"%s\" has now caught up with primary",
application_name)));
WalSndSetState(WALSNDSTATE_STREAMING);
}
/*
* When SIGUSR2 arrives, we send any outstanding logs up to the
* shutdown checkpoint record (i.e., the latest record) and exit.
* This may be a normal termination at shutdown, or a promotion,
* the walsender is not sure which.
*/
if (walsender_ready_to_stop)
{
/* ... let's just be real sure we're caught up ... */
XLogSend(output_message, &caughtup);
if (caughtup && !pq_is_send_pending())
{
walsender_shutdown_requested = true;
continue; /* don't want to wait more */
}
}
}
/*
* We don't block if not caught up, unless there is unsent data
* pending in which case we'd better block until the socket is
* write-ready. This test is only needed for the case where XLogSend
* loaded a subset of the available data but then pq_flush_if_writable
* flushed it all --- we should immediately try to send more.
*/
if (caughtup || pq_is_send_pending())
{
TimestampTz finish_time = 0;
long sleeptime = -1;
int wakeEvents;
wakeEvents = WL_LATCH_SET | WL_POSTMASTER_DEATH |
WL_SOCKET_READABLE;
if (pq_is_send_pending())
wakeEvents |= WL_SOCKET_WRITEABLE;
/* Determine time until replication timeout */
if (replication_timeout > 0)
{
long secs;
int usecs;
finish_time = TimestampTzPlusMilliseconds(last_reply_timestamp,
replication_timeout);
TimestampDifference(GetCurrentTimestamp(),
finish_time, &secs, &usecs);
sleeptime = secs * 1000 + usecs / 1000;
/* Avoid Assert in WaitLatchOrSocket if timeout is past */
if (sleeptime < 0)
sleeptime = 0;
wakeEvents |= WL_TIMEOUT;
}
/* Sleep until something happens or replication timeout */
WaitLatchOrSocket(&MyWalSnd->latch, wakeEvents,
MyProcPort->sock, sleeptime);
/*
* Check for replication timeout. Note we ignore the corner case
* possibility that the client replied just as we reached the
* timeout ... he's supposed to reply *before* that.
*/
if (replication_timeout > 0 &&
GetCurrentTimestamp() >= finish_time)
{
/*
* Since typically expiration of replication timeout means
* communication problem, we don't send the error message to
* the standby.
*/
ereport(COMMERROR,
(errmsg("terminating walsender process due to replication timeout")));
break;
}
}
}
/*
* Get here on send failure. Clean up and exit.
*
* Reset whereToSendOutput to prevent ereport from attempting to send any
* more messages to the standby.
*/
if (whereToSendOutput == DestRemote)
whereToSendOutput = DestNone;
proc_exit(0);
return 1; /* keep the compiler quiet */
}
/*
* START_REPLICATION
*/
static void
StartReplication(StartReplicationCmd *cmd)
{
StringInfoData buf;
/*
* Let postmaster know that we're streaming. Once we've declared us as a
* WAL sender process, postmaster will let us outlive the bgwriter and
* kill us last in the shutdown sequence, so we get a chance to stream all
* remaining WAL at shutdown, including the shutdown checkpoint. Note that
* there's no going back, and we mustn't write any WAL records after this.
*/
MarkPostmasterChildWalSender();
SendPostmasterSignal(PMSIGNAL_ADVANCE_STATE_MACHINE);
/*
* When promoting a cascading standby, postmaster sends SIGUSR2 to
* any cascading walsenders to kill them. But there is a corner-case where
* such walsender fails to receive SIGUSR2 and survives a standby promotion
* unexpectedly. This happens when postmaster sends SIGUSR2 before
* the walsender marks itself as a WAL sender, because postmaster sends
* SIGUSR2 to only the processes marked as a WAL sender.
*
* To avoid this corner-case, if recovery is NOT in progress even though
* the walsender is cascading one, we do the same thing as SIGUSR2 signal
* handler does, i.e., set walsender_ready_to_stop to true. Which causes
* the walsender to end later.
*
* When terminating cascading walsenders, usually postmaster writes
* the log message announcing the terminations. But there is a race condition
* here. If there is no walsender except this process before reaching here,
* postmaster thinks that there is no walsender and suppresses that
* log message. To handle this case, we always emit that log message here.
* This might cause duplicate log messages, but which is less likely to happen,
* so it's not worth writing some code to suppress them.
*/
if (am_cascading_walsender && !RecoveryInProgress())
{
ereport(LOG,
(errmsg("terminating walsender process to force cascaded standby "
"to update timeline and reconnect")));
walsender_ready_to_stop = true;
}
/*
* We assume here that we're logging enough information in the WAL for
* log-shipping, since this is checked in PostmasterMain().
*
* NOTE: wal_level can only change at shutdown, so in most cases it is
* difficult for there to be WAL data that we can still see that was written
* at wal_level='minimal'.
*/
/*
* When we first start replication the standby will be behind the primary.
* For some applications, for example, synchronous replication, it is
* important to have a clear state for this initial catchup mode, so we
* can trigger actions when we change streaming state later. We may stay
* in this state for a long time, which is exactly why we want to be able
* to monitor whether or not we are still here.
*/
WalSndSetState(WALSNDSTATE_CATCHUP);
/* Send a CopyBothResponse message, and start streaming */
pq_beginmessage(&buf, 'W');
pq_sendbyte(&buf, 0);
pq_sendint(&buf, 0, 2);
pq_endmessage(&buf);
pq_flush();
/*
* Initialize position to the received one, then the xlog records begin to
* be shipped from that position
*/
sentPtr = cmd->startpoint;
}
/* Main loop of walsender process */
static int
WalSndLoop(void)
{
char *output_message;
bool caughtup = false;
/*
* Allocate buffer that will be used for each output message. We do this
* just once to reduce palloc overhead. The buffer must be made large
* enough for maximum-sized messages.
*/
output_message = palloc(1 + sizeof(WalDataMessageHeader) + MAX_SEND_SIZE);
/*
* Allocate buffer that will be used for processing reply messages. As
* above, do this just once to reduce palloc overhead.
*/
initStringInfo(&reply_message);
/* Initialize the last reply timestamp */
last_reply_timestamp = GetCurrentTimestamp();
/* Loop forever, unless we get an error */
for (;;)
{
/*
* Emergency bailout if postmaster has died. This is to avoid the
* necessity for manual cleanup of all postmaster children.
*/
if (!PostmasterIsAlive(true))
exit(1);
/* Process any requests or signals received recently */
if (got_SIGHUP)
{
got_SIGHUP = false;
ProcessConfigFile(PGC_SIGHUP);
SyncRepInitConfig();
}
/* Normal exit from the walsender is here */
if (walsender_shutdown_requested)
{
/* Inform the standby that XLOG streaming was done */
pq_puttextmessage('C', "COPY 0");
pq_flush();
proc_exit(0);
}
/*
* If we don't have any pending data in the output buffer, try to send
* some more.
*/
if (!pq_is_send_pending())
{
XLogSend(output_message, &caughtup);
/*
* Even if we wrote all the WAL that was available when we started
* sending, more might have arrived while we were sending this
* batch. We had the latch set while sending, so we have not
* received any signals from that time. Let's arm the latch again,
* and after that check that we're still up-to-date.
*/
if (caughtup && !pq_is_send_pending())
{
ResetLatch(&MyWalSnd->latch);
XLogSend(output_message, &caughtup);
}
}
/* Flush pending output to the client */
if (pq_flush_if_writable() != 0)
break;
/*
* When SIGUSR2 arrives, we send any outstanding logs up to the
* shutdown checkpoint record (i.e., the latest record) and exit.
*/
if (walsender_ready_to_stop && !pq_is_send_pending())
{
XLogSend(output_message, &caughtup);
ProcessRepliesIfAny();
if (caughtup && !pq_is_send_pending())
walsender_shutdown_requested = true;
}
if ((caughtup || pq_is_send_pending()) &&
!got_SIGHUP &&
!walsender_shutdown_requested)
{
TimestampTz finish_time = 0;
long sleeptime;
/* Reschedule replication timeout */
if (replication_timeout > 0)
{
long secs;
int usecs;
finish_time = TimestampTzPlusMilliseconds(last_reply_timestamp,
replication_timeout);
TimestampDifference(GetCurrentTimestamp(),
finish_time, &secs, &usecs);
sleeptime = secs * 1000 + usecs / 1000;
if (WalSndDelay < sleeptime)
sleeptime = WalSndDelay;
}
else
{
/*
* XXX: Without timeout, we don't really need the periodic
* wakeups anymore, WaitLatchOrSocket should reliably wake up
* as soon as something interesting happens.
*/
sleeptime = WalSndDelay;
}
/* Sleep */
WaitLatchOrSocket(&MyWalSnd->latch, MyProcPort->sock,
true, pq_is_send_pending(),
sleeptime);
/* Check for replication timeout */
if (replication_timeout > 0 &&
GetCurrentTimestamp() >= finish_time)
{
/*
* Since typically expiration of replication timeout means
* communication problem, we don't send the error message to
* the standby.
*/
ereport(COMMERROR,
(errmsg("terminating walsender process due to replication timeout")));
break;
}
}
/*
* If we're in catchup state, see if its time to move to streaming.
* This is an important state change for users, since before this
* point data loss might occur if the primary dies and we need to
* failover to the standby. The state change is also important for
* synchronous replication, since commits that started to wait at that
* point might wait for some time.
*/
if (MyWalSnd->state == WALSNDSTATE_CATCHUP && caughtup)
{
ereport(DEBUG1,
(errmsg("standby \"%s\" has now caught up with primary",
application_name)));
WalSndSetState(WALSNDSTATE_STREAMING);
}
ProcessRepliesIfAny();
}
/*
* Get here on send failure. Clean up and exit.
*
* Reset whereToSendOutput to prevent ereport from attempting to send any
* more messages to the standby.
*/
if (whereToSendOutput == DestRemote)
whereToSendOutput = DestNone;
proc_exit(0);
return 1; /* keep the compiler quiet */
}
/*
* Read up to MAX_SEND_SIZE bytes of WAL that's been flushed to disk,
* but not yet sent to the client, and send it.
*
* msgbuf is a work area in which the output message is constructed. It's
* passed in just so we can avoid re-palloc'ing the buffer on each cycle.
* It must be of size 1 + sizeof(WalDataMessageHeader) + MAX_SEND_SIZE.
*
* If there is no unsent WAL remaining, *caughtup is set to true, otherwise
* *caughtup is set to false.
*
* Returns true if OK, false if trouble.
*/
static bool
XLogSend(char *msgbuf, bool *caughtup)
{
XLogRecPtr SendRqstPtr;
XLogRecPtr startptr;
XLogRecPtr endptr;
Size nbytes;
WalDataMessageHeader msghdr;
/*
* Attempt to send all data that's already been written out and fsync'd to
* disk. We cannot go further than what's been written out given the
* current implementation of XLogRead(). And in any case it's unsafe to
* send WAL that is not securely down to disk on the master: if the master
* subsequently crashes and restarts, slaves must not have applied any WAL
* that gets lost on the master.
*/
SendRqstPtr = GetFlushRecPtr();
/* Quick exit if nothing to do */
if (XLByteLE(SendRqstPtr, sentPtr))
{
*caughtup = true;
return true;
}
/*
* Figure out how much to send in one message. If there's no more than
* MAX_SEND_SIZE bytes to send, send everything. Otherwise send
* MAX_SEND_SIZE bytes, but round back to logfile or page boundary.
*
* The rounding is not only for performance reasons. Walreceiver relies on
* the fact that we never split a WAL record across two messages. Since a
* long WAL record is split at page boundary into continuation records,
* page boundary is always a safe cut-off point. We also assume that
* SendRqstPtr never points to the middle of a WAL record.
*/
startptr = sentPtr;
if (startptr.xrecoff >= XLogFileSize)
{
/*
* crossing a logid boundary, skip the non-existent last log segment
* in previous logical log file.
*/
startptr.xlogid += 1;
startptr.xrecoff = 0;
}
endptr = startptr;
XLByteAdvance(endptr, MAX_SEND_SIZE);
if (endptr.xlogid != startptr.xlogid)
{
/* Don't cross a logfile boundary within one message */
Assert(endptr.xlogid == startptr.xlogid + 1);
endptr.xlogid = startptr.xlogid;
endptr.xrecoff = XLogFileSize;
}
/* if we went beyond SendRqstPtr, back off */
if (XLByteLE(SendRqstPtr, endptr))
{
endptr = SendRqstPtr;
*caughtup = true;
}
else
{
/* round down to page boundary. */
endptr.xrecoff -= (endptr.xrecoff % XLOG_BLCKSZ);
*caughtup = false;
}
nbytes = endptr.xrecoff - startptr.xrecoff;
Assert(nbytes <= MAX_SEND_SIZE);
/*
* OK to read and send the slice.
*/
msgbuf[0] = 'w';
/*
* Read the log directly into the output buffer to avoid extra memcpy
* calls.
*/
XLogRead(msgbuf + 1 + sizeof(WalDataMessageHeader), startptr, nbytes);
/*
* We fill the message header last so that the send timestamp is taken as
* late as possible.
*/
msghdr.dataStart = startptr;
msghdr.walEnd = SendRqstPtr;
msghdr.sendTime = GetCurrentTimestamp();
memcpy(msgbuf + 1, &msghdr, sizeof(WalDataMessageHeader));
pq_putmessage('d', msgbuf, 1 + sizeof(WalDataMessageHeader) + nbytes);
/* Flush pending output to the client */
if (pq_flush())
return false;
sentPtr = endptr;
/* Update shared memory status */
{
/* use volatile pointer to prevent code rearrangement */
volatile WalSnd *walsnd = MyWalSnd;
SpinLockAcquire(&walsnd->mutex);
walsnd->sentPtr = sentPtr;
SpinLockRelease(&walsnd->mutex);
}
/* Report progress of XLOG streaming in PS display */
if (update_process_title)
{
char activitymsg[50];
snprintf(activitymsg, sizeof(activitymsg), "streaming %X/%X",
sentPtr.xlogid, sentPtr.xrecoff);
set_ps_display(activitymsg, false);
}
return true;
}
/*
* --------------------------------------------------------------
* ProcessIncomingNotify
*
* Deal with arriving NOTIFYs from other backends.
* This is called either directly from the SIGUSR2 signal handler,
* or the next time control reaches the outer idle loop.
* Scan pg_listener for arriving notifies, report them to my front end,
* and clear the notification field in pg_listener until next time.
*
* NOTE: since we are outside any transaction, we must create our own.
* --------------------------------------------------------------
*/
static void
ProcessIncomingNotify(void)
{
Relation lRel;
TupleDesc tdesc;
ScanKeyData key[1];
HeapScanDesc scan;
HeapTuple lTuple,
rTuple;
Datum value[Natts_pg_listener];
char repl[Natts_pg_listener],
nulls[Natts_pg_listener];
bool catchup_enabled;
/* Must prevent SIGUSR1 interrupt while I am running */
catchup_enabled = DisableCatchupInterrupt();
if (Trace_notify)
elog(DEBUG1, "ProcessIncomingNotify");
set_ps_display("notify interrupt", false);
notifyInterruptOccurred = 0;
StartTransactionCommand();
lRel = heap_open(ListenerRelationId, ExclusiveLock);
tdesc = RelationGetDescr(lRel);
/* Scan only entries with my listenerPID */
ScanKeyInit(&key[0],
Anum_pg_listener_pid,
BTEqualStrategyNumber, F_INT4EQ,
Int32GetDatum(MyProcPid));
scan = heap_beginscan(lRel, SnapshotNow, 1, key);
/* Prepare data for rewriting 0 into notification field */
nulls[0] = nulls[1] = nulls[2] = ' ';
repl[0] = repl[1] = repl[2] = ' ';
repl[Anum_pg_listener_notify - 1] = 'r';
value[0] = value[1] = value[2] = (Datum) 0;
value[Anum_pg_listener_notify - 1] = Int32GetDatum(0);
while ((lTuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{
Form_pg_listener listener = (Form_pg_listener) GETSTRUCT(lTuple);
char *relname = NameStr(listener->relname);
int32 sourcePID = listener->notification;
if (sourcePID != 0)
{
/* Notify the frontend */
if (Trace_notify)
elog(DEBUG1, "ProcessIncomingNotify: received %s from %d",
relname, (int) sourcePID);
NotifyMyFrontEnd(relname, sourcePID);
/*
* Rewrite the tuple with 0 in notification column.
*
* simple_heap_update is safe here because no one else would have
* tried to UNLISTEN us, so there can be no uncommitted changes.
*/
rTuple = heap_modifytuple(lTuple, tdesc, value, nulls, repl);
simple_heap_update(lRel, &lTuple->t_self, rTuple);
#ifdef NOT_USED /* currently there are no indexes */
CatalogUpdateIndexes(lRel, rTuple);
#endif
}
}
heap_endscan(scan);
/*
* We do NOT release the lock on pg_listener here; we need to hold it
* until end of transaction (which is about to happen, anyway) to ensure
* that other backends see our tuple updates when they look. Otherwise, a
* transaction started after this one might mistakenly think it doesn't
* need to send this backend a new NOTIFY.
*/
heap_close(lRel, NoLock);
CommitTransactionCommand();
/*
* Must flush the notify messages to ensure frontend gets them promptly.
*/
pq_flush();
set_ps_display("idle", false);
if (Trace_notify)
elog(DEBUG1, "ProcessIncomingNotify: done");
if (catchup_enabled)
EnableCatchupInterrupt();
}
static void MPPnoticeReceiver(void * arg, const PGresult * res)
{
PQExpBufferData msgbuf;
PGMessageField *pfield;
int elevel = INFO;
char * sqlstate = "00000";
char * severity = "WARNING";
char * file = "";
char * line = NULL;
char * func = "";
char message[1024];
char * detail = NULL;
char * hint = NULL;
char * context = NULL;
SegmentDatabaseDescriptor *segdbDesc = (SegmentDatabaseDescriptor *) arg;
if (!res)
return;
strcpy(message,"missing error text");
for (pfield = res->errFields; pfield != NULL; pfield = pfield->next)
{
switch (pfield->code)
{
case PG_DIAG_SEVERITY:
severity = pfield->contents;
if (strcmp(pfield->contents,"WARNING")==0)
elevel = WARNING;
else if (strcmp(pfield->contents,"NOTICE")==0)
elevel = NOTICE;
else if (strcmp(pfield->contents,"DEBUG1")==0 ||
strcmp(pfield->contents,"DEBUG")==0)
elevel = DEBUG1;
else if (strcmp(pfield->contents,"DEBUG2")==0)
elevel = DEBUG2;
else if (strcmp(pfield->contents,"DEBUG3")==0)
elevel = DEBUG3;
else if (strcmp(pfield->contents,"DEBUG4")==0)
elevel = DEBUG4;
else if (strcmp(pfield->contents,"DEBUG5")==0)
elevel = DEBUG5;
else
elevel = INFO;
break;
case PG_DIAG_SQLSTATE:
sqlstate = pfield->contents;
break;
case PG_DIAG_MESSAGE_PRIMARY:
strncpy(message, pfield->contents, 800);
message[800] = '\0';
if (segdbDesc && segdbDesc->whoami && strlen(segdbDesc->whoami) < 200)
{
strcat(message," (");
strcat(message, segdbDesc->whoami);
strcat(message,")");
}
break;
case PG_DIAG_MESSAGE_DETAIL:
detail = pfield->contents;
break;
case PG_DIAG_MESSAGE_HINT:
hint = pfield->contents;
break;
case PG_DIAG_STATEMENT_POSITION:
case PG_DIAG_INTERNAL_POSITION:
case PG_DIAG_INTERNAL_QUERY:
break;
case PG_DIAG_CONTEXT:
context = pfield->contents;
break;
case PG_DIAG_SOURCE_FILE:
file = pfield->contents;
break;
case PG_DIAG_SOURCE_LINE:
line = pfield->contents;
break;
case PG_DIAG_SOURCE_FUNCTION:
func = pfield->contents;
break;
case PG_DIAG_GP_PROCESS_TAG:
break;
default:
break;
}
}
if (elevel < client_min_messages && elevel != INFO)
return;
/*
* We use PQExpBufferData instead of StringInfoData
* because the former uses malloc, the latter palloc.
* We are in a thread, and we CANNOT use palloc since it's not
* thread safe. We cannot call elog or ereport either for the
* same reason.
*/
initPQExpBuffer(&msgbuf);
if (PG_PROTOCOL_MAJOR(FrontendProtocol) >= 3)
{
/* New style with separate fields */
appendPQExpBufferChar(&msgbuf, PG_DIAG_SEVERITY);
appendBinaryPQExpBuffer(&msgbuf, severity, strlen(severity)+1);
appendPQExpBufferChar(&msgbuf, PG_DIAG_SQLSTATE);
appendBinaryPQExpBuffer(&msgbuf, sqlstate, strlen(sqlstate)+1);
/* M field is required per protocol, so always send something */
appendPQExpBufferChar(&msgbuf, PG_DIAG_MESSAGE_PRIMARY);
appendBinaryPQExpBuffer(&msgbuf, message , strlen(message) + 1);
if (detail)
{
appendPQExpBufferChar(&msgbuf, PG_DIAG_MESSAGE_DETAIL);
appendBinaryPQExpBuffer(&msgbuf, detail, strlen(detail)+1);
}
if (hint)
{
appendPQExpBufferChar(&msgbuf, PG_DIAG_MESSAGE_HINT);
appendBinaryPQExpBuffer(&msgbuf, hint, strlen(hint)+1);
}
if (context)
{
appendPQExpBufferChar(&msgbuf, PG_DIAG_CONTEXT);
appendBinaryPQExpBuffer(&msgbuf, context, strlen(context)+1);
}
if (file)
{
appendPQExpBufferChar(&msgbuf, PG_DIAG_SOURCE_FILE);
appendBinaryPQExpBuffer(&msgbuf, file, strlen(file)+1);
}
if (line)
{
appendPQExpBufferChar(&msgbuf, PG_DIAG_SOURCE_LINE);
appendBinaryPQExpBuffer(&msgbuf, line, strlen(line)+1);
}
if (func)
{
appendPQExpBufferChar(&msgbuf, PG_DIAG_SOURCE_FUNCTION);
appendBinaryPQExpBuffer(&msgbuf, func, strlen(func)+1);
}
}
else
{
appendPQExpBuffer(&msgbuf, "%s: ", severity);
appendBinaryPQExpBuffer(&msgbuf, message, strlen(message));
appendPQExpBufferChar(&msgbuf, '\n');
appendPQExpBufferChar(&msgbuf, '\0');
}
appendPQExpBufferChar(&msgbuf, '\0'); /* terminator */
pq_putmessage('N', msgbuf.data, msgbuf.len);
termPQExpBuffer(&msgbuf);
pq_flush();
}
/*
* Receive Startup packet
* Response Client Authentication
*/
int
FileRepConnServer_ReceiveStartupPacket(void)
{
uint32 length;
int status = STATUS_OK;
char *buf = NULL;
pq_init();
status = FileRepConnServer_ReceiveMessageLength(&length);
if (status != STATUS_OK) {
goto exit;
}
if (length < (uint32) sizeof(ProtocolVersion) ||
length > MAX_STARTUP_PACKET_LENGTH) {
status = STATUS_ERROR;
ereport(WARNING,
(errcode(ERRCODE_PROTOCOL_VIOLATION),
errmsg("invalid length of startup packet"),
FileRep_errcontext()));
goto exit;
}
buf = (char *)malloc(length +1);
if (buf == NULL) {
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("not enough memory to allocate buffer for startup packet"),
FileRep_errcontext()));
}
memset(buf, 0, length + 1);
if (pq_getbytes(buf, length) == EOF) {
status = STATUS_ERROR;
ereport(WARNING,
(errcode_for_socket_access(),
errmsg("receive EOF on connection: %m"),
FileRep_errcontext()));
goto exit;
}
port->proto = ntohl(*((ProtocolVersion *) buf));
if (PG_PROTOCOL_MAJOR(port->proto) >= 3) {
/* uint32 offset = sizeof(ProtocolVersion);*/
/*
* tell the client that it is authorized (no pg_hba.conf and
* password are required).
*/
StringInfoData buf;
/* sends AUTH_REQ_OK back to client */
FakeClientAuthentication(port);
/* send to client that we are ready to receive data */
/* similar to ReadyForQuery(DestRemoteExecute); */
pq_beginmessage(&buf, 'Z');
pq_sendbyte(&buf, 'I');
pq_endmessage(&buf);
pq_flush();
} else {
ereport(WARNING,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("not supported version"),
FileRep_errcontext()));
}
exit:
if (buf) {
free(buf);
buf = NULL;
}
return status;
}
