[tarantool-patches] [v4 PATCH] replication: fix assertion with duplicate connection

[Vladimir Davydov]

On Sun, Sep 30, 2018 at 04:27:36PM +0300, Olga Arkhangelskaia wrote:
> diff --git a/src/box/box.cc b/src/box/box.cc
> index 804fc00e5..137b7ecd9 100644
> --- a/src/box/box.cc
> +++ b/src/box/box.cc
> @@ -668,7 +668,7 @@ box_sync_replication(bool connect_quorum)
>  
>  	auto guard = make_scoped_guard([=]{
>  		for (int i = 0; i < count; i++)
> -			applier_delete(appliers[i]); /* doesn't affect diag */
> +			applier_stop(appliers[i]);

Why? Now an applier leaks if an error occurs...

Is it an attempt to address my previous comments to this place?

https://www.freelists.org/post/tarantool-patches/PATCH-v2-replication-fix-assertion-with-duplicated-connect-to-same-master,1
https://www.freelists.org/post/tarantool-patches/PATCH-v3-replication-fix-assertion-with-duplicate-connection,3

What I meant is that the protocol of replicaset_connect() should be
definite: since it is passed an array of stopped appliers, it should
leave all appliers in the stopped state in case of error so that all
the caller has to do is delete them.

>  	});
>  
>  	replicaset_connect(appliers, count, connect_quorum);
> diff --git a/src/box/replication.cc b/src/box/replication.cc
> index 5755ad45e..462ab01fd 100644
> --- a/src/box/replication.cc
> +++ b/src/box/replication.cc
> @@ -271,6 +271,8 @@ replica_on_applier_connect(struct replica *replica)
>  	assert(replica->applier_sync_state == APPLIER_DISCONNECTED);
>  
>  	replica->uuid = applier->uuid;
> +	replica->applier_sync_state = APPLIER_CONNECTED;
> +	replicaset.applier.connected++;
>  
>  	struct replica *orig = replica_hash_search(&replicaset.hash, replica);
>  	if (orig != NULL && orig->applier != NULL) {
> @@ -290,6 +292,8 @@ replica_on_applier_connect(struct replica *replica)
>  
>  	if (orig != NULL) {
>  		/* Use existing struct replica */
> +		assert(orig->applier_sync_state == APPLIER_DISCONNECTED);
> +		orig->applier_sync_state = replica->applier_sync_state;
>  		replica_set_applier(orig, applier);
>  		replica_clear_applier(replica);
>  		replica_delete(replica);
> @@ -299,8 +303,6 @@ replica_on_applier_connect(struct replica *replica)
>  		replica_hash_insert(&replicaset.hash, replica);
>  	}
>  

Extra new line left.

> -	replica->applier_sync_state = APPLIER_CONNECTED;
> -	replicaset.applier.connected++;
>  }
>  
>  static void
> @@ -427,6 +429,7 @@ replicaset_update(struct applier **appliers, int count)
>  	auto uniq_guard = make_scoped_guard([&]{
>  		replica_hash_foreach_safe(&uniq, replica, next) {
>  			replica_hash_remove(&uniq, replica);
> +			replica_clear_applier(replica);
>  			replica_delete(replica);
>  		}
>  	});
> @@ -454,6 +457,8 @@ replicaset_update(struct applier **appliers, int count)
>  		replica->uuid = applier->uuid;
>  
>  		if (replica_hash_search(&uniq, replica) != NULL) {
> +			replica_clear_applier(replica);
> +			replica_delete(replica);
>  			tnt_raise(ClientError, ER_CFG, "replication",
>  				  "duplicate connection to the same replica");
>  		}
> diff --git a/test/replication/misc.test.lua b/test/replication/misc.test.lua
> index 56e1bab69..e070ebe57 100644
> --- a/test/replication/misc.test.lua
> +++ b/test/replication/misc.test.lua
> @@ -161,5 +161,47 @@ _ = test_run:wait_vclock('replica_auth', vclock)
>  test_run:cmd("stop server replica_auth")
>  test_run:cmd("cleanup server replica_auth")
>  test_run:cmd("delete server replica_auth")
> -
>  box.schema.user.drop('cluster')
> +
> +--
> +-- Test case for gh-3610. Before the fix replica would fail with the assertion
> +-- when trying to connect to the same master twice.
> +--
> +box.schema.user.grant('guest', 'replication')
> +test_run:cmd("create server replica with rpl_master=default, script='replication/replica.lua'")
> +test_run:cmd("start server replica")
> +test_run:cmd("switch replica")
> +replication = box.cfg.replication
> +box.cfg{replication = {replication, replication}}
> +
> +test_run:cmd("switch default")
> +box.schema.user.revoke('guest', 'replication')
> +test_run:cmd("stop server replica")
> +test_run:cmd('cleanup server replica')
> +test_run:cmd("delete server replica")
> +
> +-- case when replica reconnects master with duplication in new configuration
> +
> +listen = box.cfg.listen
> +box.schema.user.grant('guest', 'replication')
> +test_run:cmd("create server replica_dupl with rpl_master=default, script='replication/replica_dupl.lua'")
> +test_run:cmd("start server replica_dupl")
> +test_run:cmd("switch replica_dupl")
> +
> +test_run:cmd("switch default")
> +box.cfg{listen = ''}
> +
> +test_run:cmd("switch replica_dupl")
> +replication = box.cfg.replication
> +box.cfg{replication = {replication, replication}}
> +
> +test_run:cmd("switch default")
> +box.cfg{listen = listen}
> +fiber.sleep(1)
> +test_run:cmd("switch replica_dupl")
> +
> +test_run:cmd("switch default")
> +box.schema.user.revoke('guest', 'replication')
> +test_run:cmd("stop server replica_dupl")
> +test_run:cmd('cleanup server replica_dupl')
> +test_run:cmd("delete server replica_dupl")
> diff --git a/test/replication/replica_dupl.lua b/test/replication/replica_dupl.lua
> new file mode 100644
> index 000000000..29e94551d
> --- /dev/null
> +++ b/test/replication/replica_dupl.lua
> @@ -0,0 +1,10 @@
> +#!/usr/bin/env tarantool
> +
> +box.cfg({
> +    replication         = os.getenv("MASTER"),
> +    memtx_memory        = 107374182,
> +    replication_connect_quorum=0,
> +    replication_connect_timeout = 0.1,
> +})
> +
> +require('console').listen(os.getenv('ADMIN'))

You don't need to start another replica to test the case when a
duplicate connection is detected asynchronously: both configuration
options may be changed online.