etcd.go 9.1 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391
  1. // Copyright 2015 CoreOS, Inc.
  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 subnet
  15. import (
  16. "encoding/json"
  17. "errors"
  18. "fmt"
  19. "net"
  20. "regexp"
  21. "strconv"
  22. "time"
  23. "github.com/coreos/flannel/Godeps/_workspace/src/github.com/coreos/go-etcd/etcd"
  24. log "github.com/coreos/flannel/Godeps/_workspace/src/github.com/golang/glog"
  25. "github.com/coreos/flannel/Godeps/_workspace/src/golang.org/x/net/context"
  26. "github.com/coreos/flannel/pkg/ip"
  27. )
  28. const (
  29. registerRetries = 10
  30. subnetTTL = 24 * 3600
  31. )
  32. // etcd error codes
  33. const (
  34. etcdKeyNotFound = 100
  35. etcdKeyAlreadyExists = 105
  36. etcdEventIndexCleared = 401
  37. )
  38. type EtcdManager struct {
  39. registry Registry
  40. }
  41. var (
  42. subnetRegex *regexp.Regexp = regexp.MustCompile(`(\d+\.\d+.\d+.\d+)-(\d+)`)
  43. )
  44. type watchCursor struct {
  45. index uint64
  46. }
  47. func (c watchCursor) String() string {
  48. return strconv.FormatUint(c.index, 10)
  49. }
  50. func NewEtcdManager(config *EtcdConfig) (Manager, error) {
  51. r, err := newEtcdSubnetRegistry(config)
  52. if err != nil {
  53. return nil, err
  54. }
  55. return &EtcdManager{r}, nil
  56. }
  57. func newEtcdManager(r Registry) Manager {
  58. return &EtcdManager{r}
  59. }
  60. func (m *EtcdManager) GetNetworkConfig(ctx context.Context, network string) (*Config, error) {
  61. cfgResp, err := m.registry.getConfig(ctx, network)
  62. if err != nil {
  63. return nil, err
  64. }
  65. return ParseConfig(cfgResp.Node.Value)
  66. }
  67. func (m *EtcdManager) AcquireLease(ctx context.Context, network string, attrs *LeaseAttrs) (*Lease, error) {
  68. config, err := m.GetNetworkConfig(ctx, network)
  69. if err != nil {
  70. return nil, err
  71. }
  72. for {
  73. l, err := m.acquireLeaseOnce(ctx, network, config, attrs)
  74. switch {
  75. case err == nil:
  76. log.Info("Subnet lease acquired: ", l.Subnet)
  77. return l, nil
  78. case err == context.Canceled, err == context.DeadlineExceeded:
  79. return nil, err
  80. default:
  81. log.Error("Failed to acquire subnet: ", err)
  82. }
  83. select {
  84. case <-time.After(time.Second):
  85. case <-ctx.Done():
  86. return nil, ctx.Err()
  87. }
  88. }
  89. }
  90. func findLeaseByIP(leases []Lease, pubIP ip.IP4) *Lease {
  91. for _, l := range leases {
  92. if pubIP == l.Attrs.PublicIP {
  93. return &l
  94. }
  95. }
  96. return nil
  97. }
  98. func (m *EtcdManager) tryAcquireLease(ctx context.Context, network string, config *Config, extIP ip.IP4, attrs *LeaseAttrs) (*Lease, error) {
  99. var err error
  100. leases, _, err := m.getLeases(ctx, network)
  101. if err != nil {
  102. return nil, err
  103. }
  104. attrBytes, err := json.Marshal(attrs)
  105. if err != nil {
  106. return nil, err
  107. }
  108. // try to reuse a subnet if there's one that matches our IP
  109. if l := findLeaseByIP(leases, extIP); l != nil {
  110. resp, err := m.registry.updateSubnet(ctx, network, l.Key(), string(attrBytes), subnetTTL)
  111. if err != nil {
  112. return nil, err
  113. }
  114. l.Attrs = attrs
  115. l.Expiration = *resp.Node.Expiration
  116. return l, nil
  117. }
  118. // no existing match, grab a new one
  119. sn, err := m.allocateSubnet(config, leases)
  120. if err != nil {
  121. return nil, err
  122. }
  123. resp, err := m.registry.createSubnet(ctx, network, sn.StringSep(".", "-"), string(attrBytes), subnetTTL)
  124. switch {
  125. case err == nil:
  126. return &Lease{
  127. Subnet: sn,
  128. Attrs: attrs,
  129. Expiration: *resp.Node.Expiration,
  130. }, nil
  131. // if etcd returned Key Already Exists, try again.
  132. case err.(*etcd.EtcdError).ErrorCode == etcdKeyAlreadyExists:
  133. return nil, nil
  134. default:
  135. return nil, err
  136. }
  137. }
  138. func (m *EtcdManager) acquireLeaseOnce(ctx context.Context, network string, config *Config, attrs *LeaseAttrs) (*Lease, error) {
  139. for i := 0; i < registerRetries; i++ {
  140. l, err := m.tryAcquireLease(ctx, network, config, attrs.PublicIP, attrs)
  141. switch {
  142. case err != nil:
  143. return nil, err
  144. case l != nil:
  145. return l, nil
  146. }
  147. // before moving on, check for cancel
  148. // TODO(eyakubovich): propogate ctx deeper into registry
  149. select {
  150. case <-ctx.Done():
  151. return nil, ctx.Err()
  152. default:
  153. }
  154. }
  155. return nil, errors.New("Max retries reached trying to acquire a subnet")
  156. }
  157. func parseSubnetKey(s string) (ip.IP4Net, error) {
  158. if parts := subnetRegex.FindStringSubmatch(s); len(parts) == 3 {
  159. snIp := net.ParseIP(parts[1]).To4()
  160. prefixLen, err := strconv.ParseUint(parts[2], 10, 5)
  161. if snIp != nil && err == nil {
  162. return ip.IP4Net{IP: ip.FromIP(snIp), PrefixLen: uint(prefixLen)}, nil
  163. }
  164. }
  165. return ip.IP4Net{}, errors.New("Error parsing IP Subnet")
  166. }
  167. func (m *EtcdManager) allocateSubnet(config *Config, leases []Lease) (ip.IP4Net, error) {
  168. log.Infof("Picking subnet in range %s ... %s", config.SubnetMin, config.SubnetMax)
  169. var bag []ip.IP4
  170. sn := ip.IP4Net{IP: config.SubnetMin, PrefixLen: config.SubnetLen}
  171. OuterLoop:
  172. for ; sn.IP <= config.SubnetMax && len(bag) < 100; sn = sn.Next() {
  173. for _, l := range leases {
  174. if sn.Overlaps(l.Subnet) {
  175. continue OuterLoop
  176. }
  177. }
  178. bag = append(bag, sn.IP)
  179. }
  180. if len(bag) == 0 {
  181. return ip.IP4Net{}, errors.New("out of subnets")
  182. } else {
  183. i := randInt(0, len(bag))
  184. return ip.IP4Net{IP: bag[i], PrefixLen: config.SubnetLen}, nil
  185. }
  186. }
  187. // getLeases queries etcd to get a list of currently allocated leases for a given network.
  188. // It returns the leases along with the "as-of" etcd-index that can be used as the starting
  189. // point for etcd watch.
  190. func (m *EtcdManager) getLeases(ctx context.Context, network string) ([]Lease, uint64, error) {
  191. resp, err := m.registry.getSubnets(ctx, network)
  192. leases := []Lease{}
  193. index := uint64(0)
  194. switch {
  195. case err == nil:
  196. for _, node := range resp.Node.Nodes {
  197. sn, err := parseSubnetKey(node.Key)
  198. if err == nil {
  199. attrs := &LeaseAttrs{}
  200. if err = json.Unmarshal([]byte(node.Value), attrs); err == nil {
  201. exp := time.Time{}
  202. if node.Expiration != nil {
  203. exp = *node.Expiration
  204. }
  205. lease := Lease{
  206. Subnet: sn,
  207. Attrs: attrs,
  208. Expiration: exp,
  209. }
  210. leases = append(leases, lease)
  211. }
  212. }
  213. }
  214. index = resp.EtcdIndex
  215. case err.(*etcd.EtcdError).ErrorCode == etcdKeyNotFound:
  216. // key not found: treat it as empty set
  217. index = err.(*etcd.EtcdError).Index
  218. default:
  219. return nil, 0, err
  220. }
  221. return leases, index, nil
  222. }
  223. func (m *EtcdManager) RenewLease(ctx context.Context, network string, lease *Lease) error {
  224. attrBytes, err := json.Marshal(lease.Attrs)
  225. if err != nil {
  226. return err
  227. }
  228. // TODO(eyakubovich): propogate ctx into registry
  229. resp, err := m.registry.updateSubnet(ctx, network, lease.Key(), string(attrBytes), subnetTTL)
  230. if err != nil {
  231. return err
  232. }
  233. lease.Expiration = *resp.Node.Expiration
  234. return nil
  235. }
  236. func (m *EtcdManager) WatchLeases(ctx context.Context, network string, cursor interface{}) (WatchResult, error) {
  237. if cursor == nil {
  238. return m.watchReset(ctx, network)
  239. }
  240. nextIndex := uint64(0)
  241. if wc, ok := cursor.(watchCursor); ok {
  242. nextIndex = wc.index
  243. } else if s, ok := cursor.(string); ok {
  244. var err error
  245. nextIndex, err = strconv.ParseUint(s, 10, 64)
  246. if err != nil {
  247. return WatchResult{}, fmt.Errorf("failed to parse cursor: %v", err)
  248. }
  249. } else {
  250. return WatchResult{}, fmt.Errorf("internal error: watch cursor is of unknown type")
  251. }
  252. resp, err := m.registry.watchSubnets(ctx, network, nextIndex)
  253. switch {
  254. case err == nil:
  255. return parseSubnetWatchResponse(resp)
  256. case isIndexTooSmall(err):
  257. log.Warning("Watch of subnet leases failed because etcd index outside history window")
  258. return m.watchReset(ctx, network)
  259. default:
  260. return WatchResult{}, err
  261. }
  262. }
  263. func isIndexTooSmall(err error) bool {
  264. etcdErr, ok := err.(*etcd.EtcdError)
  265. return ok && etcdErr.ErrorCode == etcdEventIndexCleared
  266. }
  267. func parseSubnetWatchResponse(resp *etcd.Response) (WatchResult, error) {
  268. sn, err := parseSubnetKey(resp.Node.Key)
  269. if err != nil {
  270. return WatchResult{}, fmt.Errorf("error parsing subnet IP: %s", resp.Node.Key)
  271. }
  272. evt := Event{}
  273. switch resp.Action {
  274. case "delete", "expire":
  275. evt = Event{
  276. SubnetRemoved,
  277. Lease{Subnet: sn},
  278. }
  279. default:
  280. attrs := &LeaseAttrs{}
  281. err := json.Unmarshal([]byte(resp.Node.Value), attrs)
  282. if err != nil {
  283. return WatchResult{}, err
  284. }
  285. exp := time.Time{}
  286. if resp.Node.Expiration != nil {
  287. exp = *resp.Node.Expiration
  288. }
  289. evt = Event{
  290. SubnetAdded,
  291. Lease{
  292. Subnet: sn,
  293. Attrs: attrs,
  294. Expiration: exp,
  295. },
  296. }
  297. }
  298. cursor := watchCursor{resp.Node.ModifiedIndex + 1}
  299. return WatchResult{
  300. Cursor: cursor,
  301. Events: []Event{evt},
  302. }, nil
  303. }
  304. // watchReset is called when incremental watch failed and we need to grab a snapshot
  305. func (m *EtcdManager) watchReset(ctx context.Context, network string) (WatchResult, error) {
  306. wr := WatchResult{}
  307. leases, index, err := m.getLeases(ctx, network)
  308. if err != nil {
  309. return wr, fmt.Errorf("failed to retrieve subnet leases: %v", err)
  310. }
  311. cursor := watchCursor{index + 1}
  312. wr.Snapshot = leases
  313. wr.Cursor = cursor
  314. return wr, nil
  315. }
  316. func interrupted(cancel chan bool) bool {
  317. select {
  318. case <-cancel:
  319. return true
  320. default:
  321. return false
  322. }
  323. }