Re: [Tarantool-patches] [PATCH 03/12] raft: stop using replication_disconnect_timeout()

[Serge Petrenko <sergepetrenko@tarantool.org>]

LGTM.

17.11.2020 03:02, Vladislav Shpilevoy пишет:
> Raft is being moved to a separate library in src/lib. It means,
> it can't depend on anything from box/, including global
> replication parameters such as replication_timeout, and functions
> like replication_disconnect_timeout().
>
> The patch makes raft stop using replication_disconnect_timeout().
> Instead, it stores death timeout in struct raft. It is configured
> by box simultaneously with replication_timeout.
>
> Part of #5303
> ---
>   src/box/box.cc    |  2 +-
>   src/box/raftlib.c | 13 ++++++-------
>   src/box/raftlib.h | 10 +++++++---
>   3 files changed, 14 insertions(+), 11 deletions(-)
>
> diff --git a/src/box/box.cc b/src/box/box.cc
> index 8dd92a5f5..25673ed42 100644
> --- a/src/box/box.cc
> +++ b/src/box/box.cc
> @@ -890,7 +890,7 @@ void
>   box_set_replication_timeout(void)
>   {
>   	replication_timeout = box_check_replication_timeout();
> -	raft_cfg_death_timeout(box_raft());
> +	raft_cfg_death_timeout(box_raft(), replication_disconnect_timeout());
>   }
>   
>   void
> diff --git a/src/box/raftlib.c b/src/box/raftlib.c
> index 3867c63e0..c156d6f46 100644
> --- a/src/box/raftlib.c
> +++ b/src/box/raftlib.c
> @@ -804,8 +804,7 @@ raft_sm_wait_leader_dead(struct raft *raft)
>   	assert(raft->is_candidate);
>   	assert(raft->state == RAFT_STATE_FOLLOWER);
>   	assert(raft->leader != 0);
> -	double death_timeout = replication_disconnect_timeout();
> -	ev_timer_set(&raft->timer, death_timeout, death_timeout);
> +	ev_timer_set(&raft->timer, raft->death_timeout, raft->death_timeout);
>   	ev_timer_start(loop(), &raft->timer);
>   }
>   
> @@ -817,8 +816,7 @@ raft_sm_wait_leader_found(struct raft *raft)
>   	assert(raft->is_candidate);
>   	assert(raft->state == RAFT_STATE_FOLLOWER);
>   	assert(raft->leader == 0);
> -	double death_timeout = replication_disconnect_timeout();
> -	ev_timer_set(&raft->timer, death_timeout, death_timeout);
> +	ev_timer_set(&raft->timer, raft->death_timeout, raft->death_timeout);
>   	ev_timer_start(loop(), &raft->timer);
>   }
>   
> @@ -1012,14 +1010,14 @@ raft_cfg_election_quorum(struct raft *raft)
>   }
>   
>   void
> -raft_cfg_death_timeout(struct raft *raft)
> +raft_cfg_death_timeout(struct raft *raft, double death_timeout)
>   {
> +	raft->death_timeout = death_timeout;
>   	if (raft->state == RAFT_STATE_FOLLOWER && raft->is_candidate &&
>   	    raft->leader != 0) {
>   		assert(ev_is_active(&raft->timer));
> -		double death_timeout = replication_disconnect_timeout();
>   		double timeout = ev_timer_remaining(loop(), &raft->timer) -
> -				 raft->timer.at + death_timeout;
> +				 raft->timer.at + raft->death_timeout;
>   		ev_timer_stop(loop(), &raft->timer);
>   		ev_timer_set(&raft->timer, timeout, timeout);
>   		ev_timer_start(loop(), &raft->timer);
> @@ -1080,6 +1078,7 @@ raft_create(struct raft *raft)
>   		.volatile_term = 1,
>   		.term =	1,
>   		.election_timeout = 5,
> +		.death_timeout = 5,
>   	};
>   	ev_timer_init(&raft->timer, raft_sm_schedule_new_election_cb, 0, 0);
>   	raft->timer.data = raft;
> diff --git a/src/box/raftlib.h b/src/box/raftlib.h
> index 805f69d64..b33a20326 100644
> --- a/src/box/raftlib.h
> +++ b/src/box/raftlib.h
> @@ -156,6 +156,11 @@ struct raft {
>   	struct fiber *worker;
>   	/** Configured election timeout in seconds. */
>   	double election_timeout;
> +	/**
> +	 * Leader death timeout, after which it is considered dead and new
> +	 * elections can be started.
> +	 */
> +	double death_timeout;
>   	/**
>   	 * Trigger invoked each time any of the Raft node visible attributes are
>   	 * changed.
> @@ -229,11 +234,10 @@ raft_cfg_election_quorum(struct raft *raft);
>   
>   /**
>    * Configure Raft leader death timeout. I.e. number of seconds without
> - * heartbeats from the leader to consider it dead. There is no a separate
> - * option. Raft uses replication timeout for that.
> + * heartbeats from the leader to consider it dead.
>    */
>   void
> -raft_cfg_death_timeout(struct raft *raft);
> +raft_cfg_death_timeout(struct raft *raft, double death_timeout);
>   
>   /**
>    * Bump the term. When it is persisted, the node checks if there is a leader,

-- 
Serge Petrenko