-
Notifications
You must be signed in to change notification settings - Fork 379
/
serf.go
202 lines (179 loc) · 5 KB
/
serf.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
package dkron
import (
"strings"
"github.com/hashicorp/raft"
"github.com/hashicorp/serf/serf"
)
const (
// StatusReap is used to update the status of a node if we
// are handling a EventMemberReap
StatusReap = serf.MemberStatus(-1)
)
// nodeJoin is used to handle join events on the serf cluster
func (a *Agent) nodeJoin(me serf.MemberEvent) {
for _, m := range me.Members {
ok, parts := isServer(m)
if !ok {
a.logger.WithField("member", m.Name).Warn("non-server in gossip pool")
continue
}
a.logger.WithField("server", parts.Name).Info("adding server")
// Check if this server is known
found := false
a.peerLock.Lock()
existing := a.peers[parts.Region]
for idx, e := range existing {
if e.Name == parts.Name {
existing[idx] = parts
found = true
break
}
}
// Add ot the list if not known
if !found {
a.peers[parts.Region] = append(existing, parts)
}
// Check if a local peer
if parts.Region == a.config.Region {
a.localPeers[raft.ServerAddress(parts.Addr.String())] = parts
}
a.peerLock.Unlock()
// If we still expecting to bootstrap, may need to handle this
if a.config.BootstrapExpect != 0 {
a.maybeBootstrap()
}
}
}
// maybeBootstrap is used to handle bootstrapping when a new server joins
func (a *Agent) maybeBootstrap() {
// Bootstrap can only be done if there are no committed logs, remove our
// expectations of bootstrapping. This is slightly cheaper than the full
// check that BootstrapCluster will do, so this is a good pre-filter.
var index uint64
var err error
if a.raftStore != nil {
index, err = a.raftStore.LastIndex()
} else if a.raftInmem != nil {
index, err = a.raftInmem.LastIndex()
} else {
panic("neither raftInmem or raftStore is initialized")
}
if err != nil {
a.logger.WithError(err).Error("failed to read last raft index")
return
}
// Bootstrap can only be done if there are no committed logs,
// remove our expectations of bootstrapping
if index != 0 {
a.config.BootstrapExpect = 0
return
}
// Scan for all the known servers
members := a.serf.Members()
var servers []ServerParts
voters := 0
for _, member := range members {
valid, p := isServer(member)
if !valid {
continue
}
if p.Region != a.config.Region {
continue
}
if p.Expect != 0 && p.Expect != a.config.BootstrapExpect {
a.logger.WithField("member", member).Error("peer has a conflicting expect value. All nodes should expect the same number")
return
}
if p.Bootstrap {
a.logger.WithField("member", member).Error("peer has bootstrap mode. Expect disabled")
return
}
if valid {
voters++
}
servers = append(servers, *p)
}
// Skip if we haven't met the minimum expect count
if voters < a.config.BootstrapExpect {
return
}
// TODO: Query each of the servers and make sure they report no Raft peers.
// Update the peer set
// Attempt a live bootstrap!
var configuration raft.Configuration
var addrs []string
for _, server := range servers {
addr := server.Addr.String()
addrs = append(addrs, addr)
id := raft.ServerID(server.ID)
suffrage := raft.Voter
peer := raft.Server{
ID: id,
Address: raft.ServerAddress(addr),
Suffrage: suffrage,
}
configuration.Servers = append(configuration.Servers, peer)
}
a.logger.Info("agent: found expected number of peers, attempting to bootstrap cluster...",
"peers", strings.Join(addrs, ","))
future := a.raft.BootstrapCluster(configuration)
if err := future.Error(); err != nil {
a.logger.WithError(err).Error("agent: failed to bootstrap cluster")
}
// Bootstrapping complete, or failed for some reason, don't enter this again
a.config.BootstrapExpect = 0
}
// nodeFailed is used to handle fail events on the serf cluster
func (a *Agent) nodeFailed(me serf.MemberEvent) {
for _, m := range me.Members {
ok, parts := isServer(m)
if !ok {
continue
}
a.logger.Info("removing server ", parts)
// Remove the server if known
a.peerLock.Lock()
existing := a.peers[parts.Region]
n := len(existing)
for i := 0; i < n; i++ {
if existing[i].Name == parts.Name {
existing[i], existing[n-1] = existing[n-1], nil
existing = existing[:n-1]
n--
break
}
}
// Trim the list there are no known servers in a region
if n == 0 {
delete(a.peers, parts.Region)
} else {
a.peers[parts.Region] = existing
}
// Check if local peer
if parts.Region == a.config.Region {
delete(a.localPeers, raft.ServerAddress(parts.Addr.String()))
}
a.peerLock.Unlock()
}
}
// localMemberEvent is used to reconcile Serf events with the
// consistent store if we are the current leader.
func (a *Agent) localMemberEvent(me serf.MemberEvent) {
// Do nothing if we are not the leader
if !a.config.Server || !a.IsLeader() {
return
}
// Check if this is a reap event
isReap := me.EventType() == serf.EventMemberReap
// Queue the members for reconciliation
for _, m := range me.Members {
// Change the status if this is a reap event
if isReap {
m.Status = StatusReap
}
select {
case a.reconcileCh <- m:
default:
}
}
}