Commit c4de21ed authored by Willem Thiart's avatar Willem Thiart
Browse files

Fix: majority counts only include voting nodes

This fix prevents deadlocks (ie. when we are trying to get votes
from non-voting nodes that never vote).
parent 5ea572f8
......@@ -482,6 +482,10 @@ int raft_get_election_timeout(raft_server_t* me);
* @return number of nodes that this server has */
int raft_get_num_nodes(raft_server_t* me);
/**
* @return number of voting nodes that this server has */
int raft_get_num_voting_nodes(raft_server_t* me_);
/**
* @return number of items within log */
int raft_get_log_count(raft_server_t* me);
......
......@@ -275,7 +275,7 @@ int raft_recv_appendentries_response(raft_server_t* me_,
}
}
if (me->num_nodes / 2 < votes && raft_get_commit_idx(me_) < point)
if (raft_get_num_voting_nodes(me_) / 2 < votes && raft_get_commit_idx(me_) < point)
raft_set_commit_idx(me_, point);
/* Aggressively send remaining entries */
......
......@@ -51,6 +51,16 @@ int raft_get_num_nodes(raft_server_t* me_)
return ((raft_server_private_t*)me_)->num_nodes;
}
int raft_get_num_voting_nodes(raft_server_t* me_)
{
raft_server_private_t* me = (raft_server_private_t*)me_;
int i, num = 0;
for (i = 0; i < me->num_nodes; i++)
if (raft_node_is_voting(me->nodes[i]))
num++;
return num;
}
int raft_get_timeout_elapsed(raft_server_t* me_)
{
return ((raft_server_private_t*)me_)->timeout_elapsed;
......
......@@ -1981,6 +1981,57 @@ void TestRaft_leader_recv_appendentries_response_increase_commit_idx_when_majori
CuAssertIntEquals(tc, 2, raft_get_last_applied_idx(r));
}
void TestRaft_leader_recv_appendentries_response_increase_commit_idx_using_voting_nodes_majority(
CuTest * tc)
{
raft_cbs_t funcs = {
.send_appendentries = sender_appendentries,
.log = NULL
};
msg_appendentries_response_t aer;
void *sender = sender_new(NULL);
void *r = raft_new();
raft_add_node(r, NULL, 1, 1);
raft_add_node(r, NULL, 2, 0);
raft_add_node(r, NULL, 3, 0);
raft_add_non_voting_node(r, NULL, 4, 0);
raft_add_non_voting_node(r, NULL, 5, 0);
raft_set_callbacks(r, &funcs, sender);
/* I'm the leader */
raft_set_state(r, RAFT_STATE_LEADER);
raft_set_current_term(r, 1);
raft_set_commit_idx(r, 0);
/* the last applied idx will became 1, and then 2 */
raft_set_last_applied_idx(r, 0);
/* append entries - we need two */
raft_entry_t ety = {};
ety.term = 1;
ety.id = 1;
ety.data.buf = "aaaa";
ety.data.len = 4;
raft_append_entry(r, &ety);
memset(&aer, 0, sizeof(msg_appendentries_response_t));
/* FIRST entry log application */
/* send appendentries -
* server will be waiting for response */
raft_send_appendentries(r, raft_get_node(r, 2));
/* receive mock success responses */
aer.term = 1;
aer.success = 1;
aer.current_idx = 1;
aer.first_idx = 1;
raft_recv_appendentries_response(r, raft_get_node(r, 2), &aer);
CuAssertIntEquals(tc, 1, raft_get_commit_idx(r));
/* leader will now have majority followers who have appended this log */
raft_periodic(r, 1);
CuAssertIntEquals(tc, 1, raft_get_last_applied_idx(r));
}
void TestRaft_leader_recv_appendentries_response_duplicate_does_not_decrement_match_idx(
CuTest * tc)
{
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment