Re: [tarantool-patches] [PATCH v2 2/3] box: add replication_sync_timeout

[Vladimir Davydov <vdavydov.dev@gmail.com>]

On Thu, Aug 30, 2018 at 05:11:13PM +0300, Olga Arkhangelskaia wrote:
> diff --git a/src/box/lua/load_cfg.lua b/src/box/lua/load_cfg.lua
> index f803d8987..213904458 100644
> --- a/src/box/lua/load_cfg.lua
> +++ b/src/box/lua/load_cfg.lua
> @@ -200,6 +202,7 @@ local dynamic_cfg = {
>      replication_connect_timeout = private.cfg_set_replication_connect_timeout,
>      replication_connect_quorum = private.cfg_set_replication_connect_quorum,
>      replication_sync_lag    = private.cfg_set_replication_sync_lag,
> +    replication_sync_timeout    = private.cfg_set_replication_sync_timeout,

                               ^^^
Nit: three extra spaces.

>      instance_uuid           = function()
>          if box.cfg.instance_uuid ~= box.info.uuid then
>              box.error(box.error.CFG, 'instance_uuid',
> diff --git a/src/box/replication.cc b/src/box/replication.cc
> index 861ce34ea..be58b0225 100644
> --- a/src/box/replication.cc
> +++ b/src/box/replication.cc
> @@ -49,7 +49,7 @@ double replication_timeout = 1.0; /* seconds */
>  double replication_connect_timeout = 30.0; /* seconds */
>  int replication_connect_quorum = REPLICATION_CONNECT_QUORUM_ALL;
>  double replication_sync_lag = 10.0; /* seconds */
> -
> +double replication_sync_timeout = 300.0; /* seconds */
>  struct replicaset replicaset;

Nit: an empty line was here for a reason - it separated configuration
options from replicaset definition. Please leave it.

>  
>  static int
> @@ -673,12 +673,20 @@ replicaset_sync(void)
>  
>  	/*
>  	 * Wait until all connected replicas synchronize up to
> -	 * replication_sync_lag
> +	 * replication_sync_lag or return on replication_sync_timeout
>  	 */
> +	double start_time = ev_monotonic_now(loop());
> +	double deadline = start_time + replication_sync_timeout;
>  	while (replicaset.applier.synced < quorum &&
>  	       replicaset.applier.connected +
> -	       replicaset.applier.loading >= quorum)
> -		fiber_cond_wait(&replicaset.applier.cond);
> +	       replicaset.applier.loading >= quorum) {
> +		if (fiber_cond_wait_deadline(&replicaset.applier.cond,
> +				            deadline) != 0) {
> +			say_crit("replication_sync_timeout fired, entering orphan mode");

This message is not needed, because "entering orphan mode" is printed
right below in this case.

> +			break;
> +		}
> +
> +	}
>  
>  	if (replicaset.applier.synced < quorum) {
>  		/*