[tarantool-patches] [PATCH] replication: do not abort replication on ER_UNKNOWN_REPLICA

[tarantool-patches] [PATCH] replication: do not abort replication on ER_UNKNOWN_REPLICA

[Konstantin Osipov]

From: Konstantin Osipov <kostja.osipov@gmail.com>

In 3-node replica-set, registering with the leader node
does not guarantee the registration record arrives to the second
peer immediately. The joining node may bootstrap faster than
the registration record arrives to the second peer, in which
case replication will fail to create a full mesh.

Fixes gh-4455
---
 src/box/applier.cc | 1 +
 1 file changed, 1 insertion(+)

diff --git a/src/box/applier.cc b/src/box/applier.cc
index 4304ff055..6239fcfd3 100644
--- a/src/box/applier.cc
+++ b/src/box/applier.cc
@@ -100,6 +100,7 @@ applier_log_error(struct applier *applier, struct error *e)
 	case ER_ACCESS_DENIED:
 	case ER_NO_SUCH_USER:
 	case ER_SYSTEM:
+	case ER_UNKNOWN_REPLICA:
 		say_info("will retry every %.2lf second",
 			 replication_reconnect_interval());
 		break;
-- 
2.20.1

Re: [tarantool-patches] [PATCH] replication: do not abort replication on ER_UNKNOWN_REPLICA

[Vladimir Davydov]

On Tue, Aug 27, 2019 at 07:07:20PM +0300, Konstantin Osipov wrote:
> From: Konstantin Osipov <kostja.osipov@gmail.com>
> 
> In 3-node replica-set, registering with the leader node
> does not guarantee the registration record arrives to the second
> peer immediately. The joining node may bootstrap faster than
> the registration record arrives to the second peer, in which
> case replication will fail to create a full mesh.

Makes sense. Ack.

> 
> Fixes gh-4455
> ---
>  src/box/applier.cc | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/src/box/applier.cc b/src/box/applier.cc
> index 4304ff055..6239fcfd3 100644
> --- a/src/box/applier.cc
> +++ b/src/box/applier.cc
> @@ -100,6 +100,7 @@ applier_log_error(struct applier *applier, struct error *e)
>  	case ER_ACCESS_DENIED:
>  	case ER_NO_SUCH_USER:
>  	case ER_SYSTEM:
> +	case ER_UNKNOWN_REPLICA:
>  		say_info("will retry every %.2lf second",
>  			 replication_reconnect_interval());
>  		break;