[PATCH] replication: update replica gc state on subscribe

[PATCH] replication: update replica gc state on subscribe

[Vladimir Davydov]

We advance replica->gc state only when an xlog file is fully recovered,
see recovery_close_log and relay_on_close_log_f. It may turn out that an
xlog file is fully recovered, but isn't closed properly by relay (i.e.
recovery_close_log isn't called), because the replica closes connection
for some reason (e.g. timeout). If this happens, the old xlog file
won't be removed when the replica reconnects, because we don't advance
replica->gc state on reconnect, so the useless xlog file won't be
removed until the next xlog file is relayed. This results in occasional
replication/gc.test.lua failures. Fix this by updating replica->gc on
reconnect with the current replica vclock.

Closes #4034
---
https://github.com/tarantool/tarantool/issues/4034
https://github.com/tarantool/tarantool/commits/dv/gh-4034-replication-update-gc-state-on-subscribe

 src/box/relay.cc | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/src/box/relay.cc b/src/box/relay.cc
index 90fced24..945c01a5 100644
--- a/src/box/relay.cc
+++ b/src/box/relay.cc
@@ -637,13 +637,16 @@ relay_subscribe(struct replica *replica, int fd, uint64_t sync,
 	/*
 	 * Register the replica with the garbage collector
 	 * unless it has already been registered by initial
-	 * join.
+	 * join or subscribe. Otherwise update the consumer
+	 * state with the current replica vclock.
 	 */
 	if (replica->gc == NULL) {
 		replica->gc = gc_consumer_register(replica_clock, "replica %s",
 						   tt_uuid_str(&replica->uuid));
 		if (replica->gc == NULL)
 			diag_raise();
+	} else {
+		gc_consumer_advance(replica->gc, replica_clock);
 	}
 
 	relay_start(relay, fd, sync, relay_send_row);
-- 
2.11.0

Re: [PATCH] replication: update replica gc state on subscribe

[Vladimir Davydov]

On Wed, Mar 13, 2019 at 04:13:45PM +0300, Vladimir Davydov wrote:
> We advance replica->gc state only when an xlog file is fully recovered,
> see recovery_close_log and relay_on_close_log_f. It may turn out that an
> xlog file is fully recovered, but isn't closed properly by relay (i.e.
> recovery_close_log isn't called), because the replica closes connection
> for some reason (e.g. timeout). If this happens, the old xlog file
> won't be removed when the replica reconnects, because we don't advance
> replica->gc state on reconnect, so the useless xlog file won't be
> removed until the next xlog file is relayed. This results in occasional
> replication/gc.test.lua failures. Fix this by updating replica->gc on
> reconnect with the current replica vclock.
> 
> Closes #4034
> ---
> https://github.com/tarantool/tarantool/issues/4034
> https://github.com/tarantool/tarantool/commits/dv/gh-4034-replication-update-gc-state-on-subscribe
> 
>  src/box/relay.cc | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)

Alexander Tikhonov confirms that this patch fixes the test failure.
Pushed to 2.1.