Zack Williams | e940c7a | 2019-08-21 14:25:39 -0700 | [diff] [blame] | 1 | /* |
| 2 | * |
| 3 | * Copyright 2017 gRPC authors. |
| 4 | * |
| 5 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 6 | * you may not use this file except in compliance with the License. |
| 7 | * You may obtain a copy of the License at |
| 8 | * |
| 9 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 10 | * |
| 11 | * Unless required by applicable law or agreed to in writing, software |
| 12 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 13 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 14 | * See the License for the specific language governing permissions and |
| 15 | * limitations under the License. |
| 16 | * |
| 17 | */ |
| 18 | |
| 19 | package grpc |
| 20 | |
| 21 | import ( |
| 22 | "fmt" |
| 23 | "sync" |
| 24 | |
| 25 | "google.golang.org/grpc/balancer" |
| 26 | "google.golang.org/grpc/connectivity" |
| 27 | "google.golang.org/grpc/grpclog" |
| 28 | "google.golang.org/grpc/resolver" |
| 29 | ) |
| 30 | |
| 31 | // scStateUpdate contains the subConn and the new state it changed to. |
| 32 | type scStateUpdate struct { |
| 33 | sc balancer.SubConn |
| 34 | state connectivity.State |
| 35 | } |
| 36 | |
| 37 | // scStateUpdateBuffer is an unbounded channel for scStateChangeTuple. |
| 38 | // TODO make a general purpose buffer that uses interface{}. |
| 39 | type scStateUpdateBuffer struct { |
| 40 | c chan *scStateUpdate |
| 41 | mu sync.Mutex |
| 42 | backlog []*scStateUpdate |
| 43 | } |
| 44 | |
| 45 | func newSCStateUpdateBuffer() *scStateUpdateBuffer { |
| 46 | return &scStateUpdateBuffer{ |
| 47 | c: make(chan *scStateUpdate, 1), |
| 48 | } |
| 49 | } |
| 50 | |
| 51 | func (b *scStateUpdateBuffer) put(t *scStateUpdate) { |
| 52 | b.mu.Lock() |
| 53 | defer b.mu.Unlock() |
| 54 | if len(b.backlog) == 0 { |
| 55 | select { |
| 56 | case b.c <- t: |
| 57 | return |
| 58 | default: |
| 59 | } |
| 60 | } |
| 61 | b.backlog = append(b.backlog, t) |
| 62 | } |
| 63 | |
| 64 | func (b *scStateUpdateBuffer) load() { |
| 65 | b.mu.Lock() |
| 66 | defer b.mu.Unlock() |
| 67 | if len(b.backlog) > 0 { |
| 68 | select { |
| 69 | case b.c <- b.backlog[0]: |
| 70 | b.backlog[0] = nil |
| 71 | b.backlog = b.backlog[1:] |
| 72 | default: |
| 73 | } |
| 74 | } |
| 75 | } |
| 76 | |
| 77 | // get returns the channel that the scStateUpdate will be sent to. |
| 78 | // |
| 79 | // Upon receiving, the caller should call load to send another |
| 80 | // scStateChangeTuple onto the channel if there is any. |
| 81 | func (b *scStateUpdateBuffer) get() <-chan *scStateUpdate { |
| 82 | return b.c |
| 83 | } |
| 84 | |
| 85 | // ccBalancerWrapper is a wrapper on top of cc for balancers. |
| 86 | // It implements balancer.ClientConn interface. |
| 87 | type ccBalancerWrapper struct { |
| 88 | cc *ClientConn |
| 89 | balancer balancer.Balancer |
| 90 | stateChangeQueue *scStateUpdateBuffer |
| 91 | resolverUpdateCh chan *resolver.State |
| 92 | done chan struct{} |
| 93 | |
| 94 | mu sync.Mutex |
| 95 | subConns map[*acBalancerWrapper]struct{} |
| 96 | } |
| 97 | |
| 98 | func newCCBalancerWrapper(cc *ClientConn, b balancer.Builder, bopts balancer.BuildOptions) *ccBalancerWrapper { |
| 99 | ccb := &ccBalancerWrapper{ |
| 100 | cc: cc, |
| 101 | stateChangeQueue: newSCStateUpdateBuffer(), |
| 102 | resolverUpdateCh: make(chan *resolver.State, 1), |
| 103 | done: make(chan struct{}), |
| 104 | subConns: make(map[*acBalancerWrapper]struct{}), |
| 105 | } |
| 106 | go ccb.watcher() |
| 107 | ccb.balancer = b.Build(ccb, bopts) |
| 108 | return ccb |
| 109 | } |
| 110 | |
| 111 | // watcher balancer functions sequentially, so the balancer can be implemented |
| 112 | // lock-free. |
| 113 | func (ccb *ccBalancerWrapper) watcher() { |
| 114 | for { |
| 115 | select { |
| 116 | case t := <-ccb.stateChangeQueue.get(): |
| 117 | ccb.stateChangeQueue.load() |
| 118 | select { |
| 119 | case <-ccb.done: |
| 120 | ccb.balancer.Close() |
| 121 | return |
| 122 | default: |
| 123 | } |
| 124 | if ub, ok := ccb.balancer.(balancer.V2Balancer); ok { |
| 125 | ub.UpdateSubConnState(t.sc, balancer.SubConnState{ConnectivityState: t.state}) |
| 126 | } else { |
| 127 | ccb.balancer.HandleSubConnStateChange(t.sc, t.state) |
| 128 | } |
| 129 | case s := <-ccb.resolverUpdateCh: |
| 130 | select { |
| 131 | case <-ccb.done: |
| 132 | ccb.balancer.Close() |
| 133 | return |
| 134 | default: |
| 135 | } |
| 136 | if ub, ok := ccb.balancer.(balancer.V2Balancer); ok { |
| 137 | ub.UpdateResolverState(*s) |
| 138 | } else { |
| 139 | ccb.balancer.HandleResolvedAddrs(s.Addresses, nil) |
| 140 | } |
| 141 | case <-ccb.done: |
| 142 | } |
| 143 | |
| 144 | select { |
| 145 | case <-ccb.done: |
| 146 | ccb.balancer.Close() |
| 147 | ccb.mu.Lock() |
| 148 | scs := ccb.subConns |
| 149 | ccb.subConns = nil |
| 150 | ccb.mu.Unlock() |
| 151 | for acbw := range scs { |
| 152 | ccb.cc.removeAddrConn(acbw.getAddrConn(), errConnDrain) |
| 153 | } |
| 154 | return |
| 155 | default: |
| 156 | } |
| 157 | } |
| 158 | } |
| 159 | |
| 160 | func (ccb *ccBalancerWrapper) close() { |
| 161 | close(ccb.done) |
| 162 | } |
| 163 | |
| 164 | func (ccb *ccBalancerWrapper) handleSubConnStateChange(sc balancer.SubConn, s connectivity.State) { |
| 165 | // When updating addresses for a SubConn, if the address in use is not in |
| 166 | // the new addresses, the old ac will be tearDown() and a new ac will be |
| 167 | // created. tearDown() generates a state change with Shutdown state, we |
| 168 | // don't want the balancer to receive this state change. So before |
| 169 | // tearDown() on the old ac, ac.acbw (acWrapper) will be set to nil, and |
| 170 | // this function will be called with (nil, Shutdown). We don't need to call |
| 171 | // balancer method in this case. |
| 172 | if sc == nil { |
| 173 | return |
| 174 | } |
| 175 | ccb.stateChangeQueue.put(&scStateUpdate{ |
| 176 | sc: sc, |
| 177 | state: s, |
| 178 | }) |
| 179 | } |
| 180 | |
| 181 | func (ccb *ccBalancerWrapper) updateResolverState(s resolver.State) { |
| 182 | if ccb.cc.curBalancerName != grpclbName { |
| 183 | // Filter any grpclb addresses since we don't have the grpclb balancer. |
| 184 | for i := 0; i < len(s.Addresses); { |
| 185 | if s.Addresses[i].Type == resolver.GRPCLB { |
| 186 | copy(s.Addresses[i:], s.Addresses[i+1:]) |
| 187 | s.Addresses = s.Addresses[:len(s.Addresses)-1] |
| 188 | continue |
| 189 | } |
| 190 | i++ |
| 191 | } |
| 192 | } |
| 193 | select { |
| 194 | case <-ccb.resolverUpdateCh: |
| 195 | default: |
| 196 | } |
| 197 | ccb.resolverUpdateCh <- &s |
| 198 | } |
| 199 | |
| 200 | func (ccb *ccBalancerWrapper) NewSubConn(addrs []resolver.Address, opts balancer.NewSubConnOptions) (balancer.SubConn, error) { |
| 201 | if len(addrs) <= 0 { |
| 202 | return nil, fmt.Errorf("grpc: cannot create SubConn with empty address list") |
| 203 | } |
| 204 | ccb.mu.Lock() |
| 205 | defer ccb.mu.Unlock() |
| 206 | if ccb.subConns == nil { |
| 207 | return nil, fmt.Errorf("grpc: ClientConn balancer wrapper was closed") |
| 208 | } |
| 209 | ac, err := ccb.cc.newAddrConn(addrs, opts) |
| 210 | if err != nil { |
| 211 | return nil, err |
| 212 | } |
| 213 | acbw := &acBalancerWrapper{ac: ac} |
| 214 | acbw.ac.mu.Lock() |
| 215 | ac.acbw = acbw |
| 216 | acbw.ac.mu.Unlock() |
| 217 | ccb.subConns[acbw] = struct{}{} |
| 218 | return acbw, nil |
| 219 | } |
| 220 | |
| 221 | func (ccb *ccBalancerWrapper) RemoveSubConn(sc balancer.SubConn) { |
| 222 | acbw, ok := sc.(*acBalancerWrapper) |
| 223 | if !ok { |
| 224 | return |
| 225 | } |
| 226 | ccb.mu.Lock() |
| 227 | defer ccb.mu.Unlock() |
| 228 | if ccb.subConns == nil { |
| 229 | return |
| 230 | } |
| 231 | delete(ccb.subConns, acbw) |
| 232 | ccb.cc.removeAddrConn(acbw.getAddrConn(), errConnDrain) |
| 233 | } |
| 234 | |
| 235 | func (ccb *ccBalancerWrapper) UpdateBalancerState(s connectivity.State, p balancer.Picker) { |
| 236 | ccb.mu.Lock() |
| 237 | defer ccb.mu.Unlock() |
| 238 | if ccb.subConns == nil { |
| 239 | return |
| 240 | } |
| 241 | // Update picker before updating state. Even though the ordering here does |
| 242 | // not matter, it can lead to multiple calls of Pick in the common start-up |
| 243 | // case where we wait for ready and then perform an RPC. If the picker is |
| 244 | // updated later, we could call the "connecting" picker when the state is |
| 245 | // updated, and then call the "ready" picker after the picker gets updated. |
| 246 | ccb.cc.blockingpicker.updatePicker(p) |
| 247 | ccb.cc.csMgr.updateState(s) |
| 248 | } |
| 249 | |
| 250 | func (ccb *ccBalancerWrapper) ResolveNow(o resolver.ResolveNowOption) { |
| 251 | ccb.cc.resolveNow(o) |
| 252 | } |
| 253 | |
| 254 | func (ccb *ccBalancerWrapper) Target() string { |
| 255 | return ccb.cc.target |
| 256 | } |
| 257 | |
| 258 | // acBalancerWrapper is a wrapper on top of ac for balancers. |
| 259 | // It implements balancer.SubConn interface. |
| 260 | type acBalancerWrapper struct { |
| 261 | mu sync.Mutex |
| 262 | ac *addrConn |
| 263 | } |
| 264 | |
| 265 | func (acbw *acBalancerWrapper) UpdateAddresses(addrs []resolver.Address) { |
| 266 | acbw.mu.Lock() |
| 267 | defer acbw.mu.Unlock() |
| 268 | if len(addrs) <= 0 { |
| 269 | acbw.ac.tearDown(errConnDrain) |
| 270 | return |
| 271 | } |
| 272 | if !acbw.ac.tryUpdateAddrs(addrs) { |
| 273 | cc := acbw.ac.cc |
| 274 | opts := acbw.ac.scopts |
| 275 | acbw.ac.mu.Lock() |
| 276 | // Set old ac.acbw to nil so the Shutdown state update will be ignored |
| 277 | // by balancer. |
| 278 | // |
| 279 | // TODO(bar) the state transition could be wrong when tearDown() old ac |
| 280 | // and creating new ac, fix the transition. |
| 281 | acbw.ac.acbw = nil |
| 282 | acbw.ac.mu.Unlock() |
| 283 | acState := acbw.ac.getState() |
| 284 | acbw.ac.tearDown(errConnDrain) |
| 285 | |
| 286 | if acState == connectivity.Shutdown { |
| 287 | return |
| 288 | } |
| 289 | |
| 290 | ac, err := cc.newAddrConn(addrs, opts) |
| 291 | if err != nil { |
| 292 | grpclog.Warningf("acBalancerWrapper: UpdateAddresses: failed to newAddrConn: %v", err) |
| 293 | return |
| 294 | } |
| 295 | acbw.ac = ac |
| 296 | ac.mu.Lock() |
| 297 | ac.acbw = acbw |
| 298 | ac.mu.Unlock() |
| 299 | if acState != connectivity.Idle { |
| 300 | ac.connect() |
| 301 | } |
| 302 | } |
| 303 | } |
| 304 | |
| 305 | func (acbw *acBalancerWrapper) Connect() { |
| 306 | acbw.mu.Lock() |
| 307 | defer acbw.mu.Unlock() |
| 308 | acbw.ac.connect() |
| 309 | } |
| 310 | |
| 311 | func (acbw *acBalancerWrapper) getAddrConn() *addrConn { |
| 312 | acbw.mu.Lock() |
| 313 | defer acbw.mu.Unlock() |
| 314 | return acbw.ac |
| 315 | } |