local_manager.go 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499
  1. // Copyright 2015 flannel authors
  2. //
  3. // Licensed under the Apache License, Version 2.0 (the "License");
  4. // you may not use this file except in compliance with the License.
  5. // You may obtain a copy of the License at
  6. //
  7. // http://www.apache.org/licenses/LICENSE-2.0
  8. //
  9. // Unless required by applicable law or agreed to in writing, software
  10. // distributed under the License is distributed on an "AS IS" BASIS,
  11. // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  12. // See the License for the specific language governing permissions and
  13. // limitations under the License.
  14. package etcdv2
  15. import (
  16. "errors"
  17. "fmt"
  18. "strconv"
  19. "time"
  20. etcd "github.com/coreos/etcd/client"
  21. "github.com/coreos/flannel/pkg/ip"
  22. . "github.com/coreos/flannel/subnet"
  23. log "github.com/golang/glog"
  24. "golang.org/x/net/context"
  25. )
  26. const (
  27. raceRetries = 10
  28. subnetTTL = 24 * time.Hour
  29. )
  30. type LocalManager struct {
  31. registry Registry
  32. }
  33. type watchCursor struct {
  34. index uint64
  35. }
  36. func isErrEtcdTestFailed(e error) bool {
  37. if e == nil {
  38. return false
  39. }
  40. etcdErr, ok := e.(etcd.Error)
  41. return ok && etcdErr.Code == etcd.ErrorCodeTestFailed
  42. }
  43. func isErrEtcdNodeExist(e error) bool {
  44. if e == nil {
  45. return false
  46. }
  47. etcdErr, ok := e.(etcd.Error)
  48. return ok || etcdErr.Code == etcd.ErrorCodeNodeExist
  49. }
  50. func isErrEtcdKeyNotFound(e error) bool {
  51. if e == nil {
  52. return false
  53. }
  54. etcdErr, ok := e.(etcd.Error)
  55. return ok || etcdErr.Code == etcd.ErrorCodeKeyNotFound
  56. }
  57. func (c watchCursor) String() string {
  58. return strconv.FormatUint(c.index, 10)
  59. }
  60. func NewLocalManager(config *EtcdConfig) (Manager, error) {
  61. r, err := newEtcdSubnetRegistry(config, nil)
  62. if err != nil {
  63. return nil, err
  64. }
  65. return newLocalManager(r), nil
  66. }
  67. func newLocalManager(r Registry) Manager {
  68. return &LocalManager{
  69. registry: r,
  70. }
  71. }
  72. func (m *LocalManager) GetNetworkConfig(ctx context.Context, network string) (*Config, error) {
  73. cfg, err := m.registry.getNetworkConfig(ctx, network)
  74. if err != nil {
  75. return nil, err
  76. }
  77. return ParseConfig(cfg)
  78. }
  79. func (m *LocalManager) AcquireLease(ctx context.Context, network string, attrs *LeaseAttrs) (*Lease, error) {
  80. config, err := m.GetNetworkConfig(ctx, network)
  81. if err != nil {
  82. return nil, err
  83. }
  84. for i := 0; i < raceRetries; i++ {
  85. l, err := m.tryAcquireLease(ctx, network, config, attrs.PublicIP, attrs)
  86. switch err {
  87. case nil:
  88. return l, nil
  89. case errTryAgain:
  90. continue
  91. default:
  92. return nil, err
  93. }
  94. }
  95. return nil, errors.New("Max retries reached trying to acquire a subnet")
  96. }
  97. func findLeaseByIP(leases []Lease, pubIP ip.IP4) *Lease {
  98. for _, l := range leases {
  99. if pubIP == l.Attrs.PublicIP {
  100. return &l
  101. }
  102. }
  103. return nil
  104. }
  105. func (m *LocalManager) tryAcquireLease(ctx context.Context, network string, config *Config, extIaddr ip.IP4, attrs *LeaseAttrs) (*Lease, error) {
  106. leases, _, err := m.registry.getSubnets(ctx, network)
  107. if err != nil {
  108. return nil, err
  109. }
  110. // try to reuse a subnet if there's one that matches our IP
  111. if l := findLeaseByIP(leases, extIaddr); l != nil {
  112. // make sure the existing subnet is still within the configured network
  113. if isSubnetConfigCompat(config, l.Subnet) {
  114. log.Infof("Found lease (%v) for current IP (%v), reusing", l.Subnet, extIaddr)
  115. ttl := time.Duration(0)
  116. if !l.Expiration.IsZero() {
  117. // Not a reservation
  118. ttl = subnetTTL
  119. }
  120. exp, err := m.registry.updateSubnet(ctx, network, l.Subnet, attrs, ttl, 0)
  121. if err != nil {
  122. return nil, err
  123. }
  124. l.Attrs = *attrs
  125. l.Expiration = exp
  126. return l, nil
  127. } else {
  128. log.Infof("Found lease (%v) for current IP (%v) but not compatible with current config, deleting", l.Subnet, extIaddr)
  129. if err := m.registry.deleteSubnet(ctx, network, l.Subnet); err != nil {
  130. return nil, err
  131. }
  132. }
  133. }
  134. // no existing match, grab a new one
  135. sn, err := m.allocateSubnet(config, leases)
  136. if err != nil {
  137. return nil, err
  138. }
  139. exp, err := m.registry.createSubnet(ctx, network, sn, attrs, subnetTTL)
  140. switch {
  141. case err == nil:
  142. return &Lease{
  143. Subnet: sn,
  144. Attrs: *attrs,
  145. Expiration: exp,
  146. }, nil
  147. case isErrEtcdNodeExist(err):
  148. return nil, errTryAgain
  149. default:
  150. return nil, err
  151. }
  152. }
  153. func (m *LocalManager) allocateSubnet(config *Config, leases []Lease) (ip.IP4Net, error) {
  154. log.Infof("Picking subnet in range %s ... %s", config.SubnetMin, config.SubnetMax)
  155. var bag []ip.IP4
  156. sn := ip.IP4Net{IP: config.SubnetMin, PrefixLen: config.SubnetLen}
  157. OuterLoop:
  158. for ; sn.IP <= config.SubnetMax && len(bag) < 100; sn = sn.Next() {
  159. for _, l := range leases {
  160. if sn.Overlaps(l.Subnet) {
  161. continue OuterLoop
  162. }
  163. }
  164. bag = append(bag, sn.IP)
  165. }
  166. if len(bag) == 0 {
  167. return ip.IP4Net{}, errors.New("out of subnets")
  168. } else {
  169. i := randInt(0, len(bag))
  170. return ip.IP4Net{IP: bag[i], PrefixLen: config.SubnetLen}, nil
  171. }
  172. }
  173. func (m *LocalManager) RevokeLease(ctx context.Context, network string, sn ip.IP4Net) error {
  174. return m.registry.deleteSubnet(ctx, network, sn)
  175. }
  176. func (m *LocalManager) RenewLease(ctx context.Context, network string, lease *Lease) error {
  177. exp, err := m.registry.updateSubnet(ctx, network, lease.Subnet, &lease.Attrs, subnetTTL, 0)
  178. if err != nil {
  179. return err
  180. }
  181. lease.Expiration = exp
  182. return nil
  183. }
  184. func getNextIndex(cursor interface{}) (uint64, error) {
  185. nextIndex := uint64(0)
  186. if wc, ok := cursor.(watchCursor); ok {
  187. nextIndex = wc.index
  188. } else if s, ok := cursor.(string); ok {
  189. var err error
  190. nextIndex, err = strconv.ParseUint(s, 10, 64)
  191. if err != nil {
  192. return 0, fmt.Errorf("failed to parse cursor: %v", err)
  193. }
  194. } else {
  195. return 0, fmt.Errorf("internal error: watch cursor is of unknown type")
  196. }
  197. return nextIndex, nil
  198. }
  199. func (m *LocalManager) leaseWatchReset(ctx context.Context, network string, sn ip.IP4Net) (LeaseWatchResult, error) {
  200. l, index, err := m.registry.getSubnet(ctx, network, sn)
  201. if err != nil {
  202. return LeaseWatchResult{}, err
  203. }
  204. return LeaseWatchResult{
  205. Snapshot: []Lease{*l},
  206. Cursor: watchCursor{index},
  207. }, nil
  208. }
  209. func (m *LocalManager) WatchLease(ctx context.Context, network string, sn ip.IP4Net, cursor interface{}) (LeaseWatchResult, error) {
  210. if cursor == nil {
  211. return m.leaseWatchReset(ctx, network, sn)
  212. }
  213. nextIndex, err := getNextIndex(cursor)
  214. if err != nil {
  215. return LeaseWatchResult{}, err
  216. }
  217. evt, index, err := m.registry.watchSubnet(ctx, network, nextIndex, sn)
  218. switch {
  219. case err == nil:
  220. return LeaseWatchResult{
  221. Events: []Event{evt},
  222. Cursor: watchCursor{index},
  223. }, nil
  224. case isIndexTooSmall(err):
  225. log.Warning("Watch of subnet leases failed because etcd index outside history window")
  226. return m.leaseWatchReset(ctx, network, sn)
  227. default:
  228. return LeaseWatchResult{}, err
  229. }
  230. }
  231. func (m *LocalManager) WatchLeases(ctx context.Context, network string, cursor interface{}) (LeaseWatchResult, error) {
  232. if cursor == nil {
  233. return m.leasesWatchReset(ctx, network)
  234. }
  235. nextIndex, err := getNextIndex(cursor)
  236. if err != nil {
  237. return LeaseWatchResult{}, err
  238. }
  239. evt, index, err := m.registry.watchSubnets(ctx, network, nextIndex)
  240. switch {
  241. case err == nil:
  242. return LeaseWatchResult{
  243. Events: []Event{evt},
  244. Cursor: watchCursor{index},
  245. }, nil
  246. case isIndexTooSmall(err):
  247. log.Warning("Watch of subnet leases failed because etcd index outside history window")
  248. return m.leasesWatchReset(ctx, network)
  249. default:
  250. return LeaseWatchResult{}, err
  251. }
  252. }
  253. func (m *LocalManager) WatchNetworks(ctx context.Context, cursor interface{}) (NetworkWatchResult, error) {
  254. if cursor == nil {
  255. return m.networkWatchReset(ctx)
  256. }
  257. nextIndex, err := getNextIndex(cursor)
  258. if err != nil {
  259. return NetworkWatchResult{}, err
  260. }
  261. for {
  262. evt, index, err := m.registry.watchNetworks(ctx, nextIndex)
  263. switch {
  264. case err == nil:
  265. return NetworkWatchResult{
  266. Events: []Event{evt},
  267. Cursor: watchCursor{index},
  268. }, nil
  269. case err == errTryAgain:
  270. nextIndex = index
  271. case isIndexTooSmall(err):
  272. log.Warning("Watch of networks failed because etcd index outside history window")
  273. return m.networkWatchReset(ctx)
  274. default:
  275. return NetworkWatchResult{}, err
  276. }
  277. }
  278. }
  279. func isIndexTooSmall(err error) bool {
  280. etcdErr, ok := err.(etcd.Error)
  281. return ok && etcdErr.Code == etcd.ErrorCodeEventIndexCleared
  282. }
  283. // leasesWatchReset is called when incremental lease watch failed and we need to grab a snapshot
  284. func (m *LocalManager) leasesWatchReset(ctx context.Context, network string) (LeaseWatchResult, error) {
  285. wr := LeaseWatchResult{}
  286. leases, index, err := m.registry.getSubnets(ctx, network)
  287. if err != nil {
  288. return wr, fmt.Errorf("failed to retrieve subnet leases: %v", err)
  289. }
  290. wr.Cursor = watchCursor{index}
  291. wr.Snapshot = leases
  292. return wr, nil
  293. }
  294. // networkWatchReset is called when incremental network watch failed and we need to grab a snapshot
  295. func (m *LocalManager) networkWatchReset(ctx context.Context) (NetworkWatchResult, error) {
  296. wr := NetworkWatchResult{}
  297. networks, index, err := m.registry.getNetworks(ctx)
  298. if err != nil {
  299. return wr, fmt.Errorf("failed to retrieve networks: %v", err)
  300. }
  301. wr.Cursor = watchCursor{index}
  302. wr.Snapshot = networks
  303. return wr, nil
  304. }
  305. func isSubnetConfigCompat(config *Config, sn ip.IP4Net) bool {
  306. if sn.IP < config.SubnetMin || sn.IP > config.SubnetMax {
  307. return false
  308. }
  309. return sn.PrefixLen == config.SubnetLen
  310. }
  311. func (m *LocalManager) tryAddReservation(ctx context.Context, network string, r *Reservation) error {
  312. attrs := &LeaseAttrs{
  313. PublicIP: r.PublicIP,
  314. }
  315. _, err := m.registry.createSubnet(ctx, network, r.Subnet, attrs, 0)
  316. switch {
  317. case err == nil:
  318. return nil
  319. case !isErrEtcdNodeExist(err):
  320. return err
  321. }
  322. // This subnet or its reservation already exists.
  323. // Get what's there and
  324. // - if PublicIP matches, remove the TTL make it a reservation
  325. // - otherwise, error out
  326. sub, asof, err := m.registry.getSubnet(ctx, network, r.Subnet)
  327. switch {
  328. case err == nil:
  329. case isErrEtcdKeyNotFound(err):
  330. // Subnet just got expired or was deleted
  331. return errTryAgain
  332. default:
  333. return err
  334. }
  335. if sub.Attrs.PublicIP != r.PublicIP {
  336. // Subnet already taken
  337. return ErrLeaseTaken
  338. }
  339. // remove TTL
  340. _, err = m.registry.updateSubnet(ctx, network, r.Subnet, &sub.Attrs, 0, asof)
  341. if isErrEtcdTestFailed(err) {
  342. return errTryAgain
  343. }
  344. return err
  345. }
  346. func (m *LocalManager) AddReservation(ctx context.Context, network string, r *Reservation) error {
  347. config, err := m.GetNetworkConfig(ctx, network)
  348. if err != nil {
  349. return err
  350. }
  351. if config.SubnetLen != r.Subnet.PrefixLen {
  352. return fmt.Errorf("reservation subnet has mask incompatible with network config")
  353. }
  354. if !config.Network.Overlaps(r.Subnet) {
  355. return fmt.Errorf("reservation subnet is outside of flannel network")
  356. }
  357. for i := 0; i < raceRetries; i++ {
  358. err := m.tryAddReservation(ctx, network, r)
  359. switch {
  360. case err == nil:
  361. return nil
  362. case err == errTryAgain:
  363. continue
  364. default:
  365. return err
  366. }
  367. }
  368. return ErrNoMoreTries
  369. }
  370. func (m *LocalManager) tryRemoveReservation(ctx context.Context, network string, subnet ip.IP4Net) error {
  371. sub, asof, err := m.registry.getSubnet(ctx, network, subnet)
  372. if err != nil {
  373. return err
  374. }
  375. // add back the TTL
  376. _, err = m.registry.updateSubnet(ctx, network, subnet, &sub.Attrs, subnetTTL, asof)
  377. if isErrEtcdTestFailed(err) {
  378. return errTryAgain
  379. }
  380. return err
  381. }
  382. //RemoveReservation removes the subnet by setting TTL back to subnetTTL (24hours)
  383. func (m *LocalManager) RemoveReservation(ctx context.Context, network string, subnet ip.IP4Net) error {
  384. for i := 0; i < raceRetries; i++ {
  385. err := m.tryRemoveReservation(ctx, network, subnet)
  386. switch {
  387. case err == nil:
  388. return nil
  389. case err == errTryAgain:
  390. continue
  391. default:
  392. return err
  393. }
  394. }
  395. return ErrNoMoreTries
  396. }
  397. func (m *LocalManager) ListReservations(ctx context.Context, network string) ([]Reservation, error) {
  398. subnets, _, err := m.registry.getSubnets(ctx, network)
  399. if err != nil {
  400. return nil, err
  401. }
  402. rsvs := []Reservation{}
  403. for _, sub := range subnets {
  404. // Reservations don't have TTL and so no expiration
  405. if !sub.Expiration.IsZero() {
  406. continue
  407. }
  408. r := Reservation{
  409. Subnet: sub.Subnet,
  410. PublicIP: sub.Attrs.PublicIP,
  411. }
  412. rsvs = append(rsvs, r)
  413. }
  414. return rsvs, nil
  415. }