postgresql/src/backend/replication
Tom Lane 1cff1b95ab Represent Lists as expansible arrays, not chains of cons-cells.
Originally, Postgres Lists were a more or less exact reimplementation of
Lisp lists, which consist of chains of separately-allocated cons cells,
each having a value and a next-cell link.  We'd hacked that once before
(commit d0b4399d8) to add a separate List header, but the data was still
in cons cells.  That makes some operations -- notably list_nth() -- O(N),
and it's bulky because of the next-cell pointers and per-cell palloc
overhead, and it's very cache-unfriendly if the cons cells end up
scattered around rather than being adjacent.

In this rewrite, we still have List headers, but the data is in a
resizable array of values, with no next-cell links.  Now we need at
most two palloc's per List, and often only one, since we can allocate
some values in the same palloc call as the List header.  (Of course,
extending an existing List may require repalloc's to enlarge the array.
But this involves just O(log N) allocations not O(N).)

Of course this is not without downsides.  The key difficulty is that
addition or deletion of a list entry may now cause other entries to
move, which it did not before.

For example, that breaks foreach() and sister macros, which historically
used a pointer to the current cons-cell as loop state.  We can repair
those macros transparently by making their actual loop state be an
integer list index; the exposed "ListCell *" pointer is no longer state
carried across loop iterations, but is just a derived value.  (In
practice, modern compilers can optimize things back to having just one
loop state value, at least for simple cases with inline loop bodies.)
In principle, this is a semantics change for cases where the loop body
inserts or deletes list entries ahead of the current loop index; but
I found no such cases in the Postgres code.

The change is not at all transparent for code that doesn't use foreach()
but chases lists "by hand" using lnext().  The largest share of such
code in the backend is in loops that were maintaining "prev" and "next"
variables in addition to the current-cell pointer, in order to delete
list cells efficiently using list_delete_cell().  However, we no longer
need a previous-cell pointer to delete a list cell efficiently.  Keeping
a next-cell pointer doesn't work, as explained above, but we can improve
matters by changing such code to use a regular foreach() loop and then
using the new macro foreach_delete_current() to delete the current cell.
(This macro knows how to update the associated foreach loop's state so
that no cells will be missed in the traversal.)

There remains a nontrivial risk of code assuming that a ListCell *
pointer will remain good over an operation that could now move the list
contents.  To help catch such errors, list.c can be compiled with a new
define symbol DEBUG_LIST_MEMORY_USAGE that forcibly moves list contents
whenever that could possibly happen.  This makes list operations
significantly more expensive so it's not normally turned on (though it
is on by default if USE_VALGRIND is on).

There are two notable API differences from the previous code:

* lnext() now requires the List's header pointer in addition to the
current cell's address.

* list_delete_cell() no longer requires a previous-cell argument.

These changes are somewhat unfortunate, but on the other hand code using
either function needs inspection to see if it is assuming anything
it shouldn't, so it's not all bad.

Programmers should be aware of these significant performance changes:

* list_nth() and related functions are now O(1); so there's no
major access-speed difference between a list and an array.

* Inserting or deleting a list element now takes time proportional to
the distance to the end of the list, due to moving the array elements.
(However, it typically *doesn't* require palloc or pfree, so except in
long lists it's probably still faster than before.)  Notably, lcons()
used to be about the same cost as lappend(), but that's no longer true
if the list is long.  Code that uses lcons() and list_delete_first()
to maintain a stack might usefully be rewritten to push and pop at the
end of the list rather than the beginning.

* There are now list_insert_nth...() and list_delete_nth...() functions
that add or remove a list cell identified by index.  These have the
data-movement penalty explained above, but there's no search penalty.

* list_concat() and variants now copy the second list's data into
storage belonging to the first list, so there is no longer any
sharing of cells between the input lists.  The second argument is
now declared "const List *" to reflect that it isn't changed.

This patch just does the minimum needed to get the new implementation
in place and fix bugs exposed by the regression tests.  As suggested
by the foregoing, there's a fair amount of followup work remaining to
do.

Also, the ENABLE_LIST_COMPAT macros are finally removed in this
commit.  Code using those should have been gone a dozen years ago.

Patch by me; thanks to David Rowley, Jesper Pedersen, and others
for review.

Discussion: https://postgr.es/m/11587.1550975080@sss.pgh.pa.us
2019-07-15 13:41:58 -04:00
..
libpqwalreceiver Phase 2 pgindent run for v12. 2019-05-22 13:04:48 -04:00
logical Use consistent style for checking return from system calls 2019-07-07 15:28:49 +02:00
pgoutput Phase 2 pgindent run for v12. 2019-05-22 13:04:48 -04:00
.gitignore Support multiple synchronous standby servers. 2016-04-06 17:18:25 +09:00
basebackup.c Represent Lists as expansible arrays, not chains of cons-cells. 2019-07-15 13:41:58 -04:00
Makefile Rethink flex flags for syncrep_scanner.l. 2017-05-19 18:05:20 -04:00
README Rename "pg_xlog" directory to "pg_wal". 2016-10-20 11:32:18 -04:00
repl_gram.y Update copyright for 2019 2019-01-02 12:44:25 -05:00
repl_scanner.l Update copyright for 2019 2019-01-02 12:44:25 -05:00
slot.c Use consistent style for checking return from system calls 2019-07-07 15:28:49 +02:00
slotfuncs.c Initial pgindent run for v12. 2019-05-22 12:55:34 -04:00
syncrep_gram.y Update copyright for 2019 2019-01-02 12:44:25 -05:00
syncrep_scanner.l Update copyright for 2019 2019-01-02 12:44:25 -05:00
syncrep.c Represent Lists as expansible arrays, not chains of cons-cells. 2019-07-15 13:41:58 -04:00
walreceiver.c Initial pgindent run for v12. 2019-05-22 12:55:34 -04:00
walreceiverfuncs.c Update copyright for 2019 2019-01-02 12:44:25 -05:00
walsender.c Use consistent style for checking return from system calls 2019-07-07 15:28:49 +02:00

src/backend/replication/README

Walreceiver - libpqwalreceiver API
----------------------------------

The transport-specific part of walreceiver, responsible for connecting to
the primary server, receiving WAL files and sending messages, is loaded
dynamically to avoid having to link the main server binary with libpq.
The dynamically loaded module is in libpqwalreceiver subdirectory.

The dynamically loaded module implements four functions:


bool walrcv_connect(char *conninfo, XLogRecPtr startpoint)

Establish connection to the primary, and starts streaming from 'startpoint'.
Returns true on success.

int walrcv_receive(char **buffer, pgsocket *wait_fd)

Retrieve any message available without blocking through the
connection.  If a message was successfully read, returns its
length. If the connection is closed, returns -1.  Otherwise returns 0
to indicate that no data is available, and sets *wait_fd to a socket
descriptor which can be waited on before trying again.  On success, a
pointer to the message payload is stored in *buffer. The returned
buffer is valid until the next call to walrcv_* functions, and the
caller should not attempt to free it.

void walrcv_send(const char *buffer, int nbytes)

Send a message to XLOG stream.

void walrcv_disconnect(void);

Disconnect.


This API should be considered internal at the moment, but we could open it
up for 3rd party replacements of libpqwalreceiver in the future, allowing
pluggable methods for receiving WAL.

Walreceiver IPC
---------------

When the WAL replay in startup process has reached the end of archived WAL,
restorable using restore_command, it starts up the walreceiver process
to fetch more WAL (if streaming replication is configured).

Walreceiver is a postmaster subprocess, so the startup process can't fork it
directly. Instead, it sends a signal to postmaster, asking postmaster to launch
it. Before that, however, startup process fills in WalRcvData->conninfo
and WalRcvData->slotname, and initializes the starting point in
WalRcvData->receiveStart.

As walreceiver receives WAL from the master server, and writes and flushes
it to disk (in pg_wal), it updates WalRcvData->receivedUpto and signals
the startup process to know how far WAL replay can advance.

Walreceiver sends information about replication progress to the master server
whenever it either writes or flushes new WAL, or the specified interval elapses.
This is used for reporting purpose.

Walsender IPC
-------------

At shutdown, postmaster handles walsender processes differently from regular
backends. It waits for regular backends to die before writing the
shutdown checkpoint and terminating pgarch and other auxiliary processes, but
that's not desirable for walsenders, because we want the standby servers to
receive all the WAL, including the shutdown checkpoint, before the master
is shut down. Therefore postmaster treats walsenders like the pgarch process,
and instructs them to terminate at PM_SHUTDOWN_2 phase, after all regular
backends have died and checkpointer has issued the shutdown checkpoint.

When postmaster accepts a connection, it immediately forks a new process
to handle the handshake and authentication, and the process initializes to
become a backend. Postmaster doesn't know if the process becomes a regular
backend or a walsender process at that time - that's indicated in the
connection handshake - so we need some extra signaling to let postmaster
identify walsender processes.

When walsender process starts up, it marks itself as a walsender process in
the PMSignal array. That way postmaster can tell it apart from regular
backends.

Note that no big harm is done if postmaster thinks that a walsender is a
regular backend; it will just terminate the walsender earlier in the shutdown
phase. A walsender will look like a regular backend until it's done with the
initialization and has marked itself in PMSignal array, and at process
termination, after unmarking the PMSignal slot.

Each walsender allocates an entry from the WalSndCtl array, and tracks
information about replication progress. User can monitor them via
statistics views.


Walsender - walreceiver protocol
--------------------------------

See manual.