Commit f246ea3b authored by Noah Misch's avatar Noah Misch

In caught-up logical walsender, sleep only in WalSndWaitForWal().

Before sleeping, WalSndWaitForWal() sends a keepalive if MyWalSnd->write
< sentPtr.  When the latest physical LSN yields no logical replication
messages (a common case), that keepalive elicits a reply.  Processing
the reply updates pg_stat_replication.replay_lsn.  WalSndLoop() lacks
that; when WalSndLoop() slept, replay_lsn advancement could stall until
wal_receiver_status_interval elapsed.  This sometimes stalled
src/test/subscription/t/001_rep_changes.pl for up to 10s.

Reviewed by Fujii Masao and Michael Paquier.

Discussion: https://postgr.es/m/20200418070142.GA1075445@rfd.leadboat.com
parent 72a3dc32
...@@ -1428,8 +1428,10 @@ WalSndWaitForWal(XLogRecPtr loc) ...@@ -1428,8 +1428,10 @@ WalSndWaitForWal(XLogRecPtr loc)
/* /*
* We only send regular messages to the client for full decoded * We only send regular messages to the client for full decoded
* transactions, but a synchronous replication and walsender shutdown * transactions, but a synchronous replication and walsender shutdown
* possibly are waiting for a later location. So we send pings * possibly are waiting for a later location. So, before sleeping, we
* containing the flush location every now and then. * send a ping containing the flush location. If the receiver is
* otherwise idle, this keepalive will trigger a reply. Processing the
* reply will update these MyWalSnd locations.
*/ */
if (MyWalSnd->flush < sentPtr && if (MyWalSnd->flush < sentPtr &&
MyWalSnd->write < sentPtr && MyWalSnd->write < sentPtr &&
...@@ -2314,14 +2316,14 @@ WalSndLoop(WalSndSendDataCallback send_data) ...@@ -2314,14 +2316,14 @@ WalSndLoop(WalSndSendDataCallback send_data)
WalSndKeepaliveIfNecessary(); WalSndKeepaliveIfNecessary();
/* /*
* We don't block if not caught up, unless there is unsent data * Block if we have unsent data. XXX For logical replication, let
* pending in which case we'd better block until the socket is * WalSndWaitForWal() handle any other blocking; idle receivers need
* write-ready. This test is only needed for the case where the * its additional actions. For physical replication, also block if
* send_data callback handled a subset of the available data but then * caught up; its send_data does not block.
* pq_flush_if_writable flushed it all --- we should immediately try
* to send more.
*/ */
if ((WalSndCaughtUp && !streamingDoneSending) || pq_is_send_pending()) if ((WalSndCaughtUp && send_data != XLogSendLogical &&
!streamingDoneSending) ||
pq_is_send_pending())
{ {
long sleeptime; long sleeptime;
int wakeEvents; int wakeEvents;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment