etcd.go 8.6 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369
  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. func NewEtcdManager(config *EtcdConfig) (Manager, error) {
  45. r, err := newEtcdSubnetRegistry(config)
  46. if err != nil {
  47. return nil, err
  48. }
  49. return &EtcdManager{r}, nil
  50. }
  51. func newEtcdManager(r Registry) Manager {
  52. return &EtcdManager{r}
  53. }
  54. func (m *EtcdManager) GetNetworkConfig(ctx context.Context, network string) (*Config, error) {
  55. cfgResp, err := m.registry.getConfig(ctx, network)
  56. if err != nil {
  57. return nil, err
  58. }
  59. return ParseConfig(cfgResp.Node.Value)
  60. }
  61. func (m *EtcdManager) AcquireLease(ctx context.Context, network string, attrs *LeaseAttrs) (*Lease, error) {
  62. config, err := m.GetNetworkConfig(ctx, network)
  63. if err != nil {
  64. return nil, err
  65. }
  66. for {
  67. l, err := m.acquireLeaseOnce(ctx, network, config, attrs)
  68. switch {
  69. case err == nil:
  70. log.Info("Subnet lease acquired: ", l.Subnet)
  71. return l, nil
  72. case err == context.Canceled, err == context.DeadlineExceeded:
  73. return nil, err
  74. default:
  75. log.Error("Failed to acquire subnet: ", err)
  76. }
  77. select {
  78. case <-time.After(time.Second):
  79. case <-ctx.Done():
  80. return nil, ctx.Err()
  81. }
  82. }
  83. }
  84. func findLeaseByIP(leases []Lease, pubIP ip.IP4) *Lease {
  85. for _, l := range leases {
  86. if pubIP == l.Attrs.PublicIP {
  87. return &l
  88. }
  89. }
  90. return nil
  91. }
  92. func (m *EtcdManager) tryAcquireLease(ctx context.Context, network string, config *Config, extIP ip.IP4, attrs *LeaseAttrs) (*Lease, error) {
  93. var err error
  94. leases, _, err := m.getLeases(ctx, network)
  95. if err != nil {
  96. return nil, err
  97. }
  98. attrBytes, err := json.Marshal(attrs)
  99. if err != nil {
  100. return nil, err
  101. }
  102. // try to reuse a subnet if there's one that matches our IP
  103. if l := findLeaseByIP(leases, extIP); l != nil {
  104. resp, err := m.registry.updateSubnet(ctx, network, l.Key(), string(attrBytes), subnetTTL)
  105. if err != nil {
  106. return nil, err
  107. }
  108. l.Attrs = attrs
  109. l.Expiration = *resp.Node.Expiration
  110. return l, nil
  111. }
  112. // no existing match, grab a new one
  113. sn, err := m.allocateSubnet(config, leases)
  114. if err != nil {
  115. return nil, err
  116. }
  117. resp, err := m.registry.createSubnet(ctx, network, sn.StringSep(".", "-"), string(attrBytes), subnetTTL)
  118. switch {
  119. case err == nil:
  120. return &Lease{
  121. Subnet: sn,
  122. Attrs: attrs,
  123. Expiration: *resp.Node.Expiration,
  124. }, nil
  125. // if etcd returned Key Already Exists, try again.
  126. case err.(*etcd.EtcdError).ErrorCode == etcdKeyAlreadyExists:
  127. return nil, nil
  128. default:
  129. return nil, err
  130. }
  131. }
  132. func (m *EtcdManager) acquireLeaseOnce(ctx context.Context, network string, config *Config, attrs *LeaseAttrs) (*Lease, error) {
  133. for i := 0; i < registerRetries; i++ {
  134. l, err := m.tryAcquireLease(ctx, network, config, attrs.PublicIP, attrs)
  135. switch {
  136. case err != nil:
  137. return nil, err
  138. case l != nil:
  139. return l, nil
  140. }
  141. // before moving on, check for cancel
  142. // TODO(eyakubovich): propogate ctx deeper into registry
  143. select {
  144. case <-ctx.Done():
  145. return nil, ctx.Err()
  146. default:
  147. }
  148. }
  149. return nil, errors.New("Max retries reached trying to acquire a subnet")
  150. }
  151. func parseSubnetKey(s string) (ip.IP4Net, error) {
  152. if parts := subnetRegex.FindStringSubmatch(s); len(parts) == 3 {
  153. snIp := net.ParseIP(parts[1]).To4()
  154. prefixLen, err := strconv.ParseUint(parts[2], 10, 5)
  155. if snIp != nil && err == nil {
  156. return ip.IP4Net{IP: ip.FromIP(snIp), PrefixLen: uint(prefixLen)}, nil
  157. }
  158. }
  159. return ip.IP4Net{}, errors.New("Error parsing IP Subnet")
  160. }
  161. func (m *EtcdManager) allocateSubnet(config *Config, leases []Lease) (ip.IP4Net, error) {
  162. log.Infof("Picking subnet in range %s ... %s", config.SubnetMin, config.SubnetMax)
  163. var bag []ip.IP4
  164. sn := ip.IP4Net{IP: config.SubnetMin, PrefixLen: config.SubnetLen}
  165. OuterLoop:
  166. for ; sn.IP <= config.SubnetMax && len(bag) < 100; sn = sn.Next() {
  167. for _, l := range leases {
  168. if sn.Overlaps(l.Subnet) {
  169. continue OuterLoop
  170. }
  171. }
  172. bag = append(bag, sn.IP)
  173. }
  174. if len(bag) == 0 {
  175. return ip.IP4Net{}, errors.New("out of subnets")
  176. } else {
  177. i := randInt(0, len(bag))
  178. return ip.IP4Net{IP: bag[i], PrefixLen: config.SubnetLen}, nil
  179. }
  180. }
  181. // getLeases queries etcd to get a list of currently allocated leases for a given network.
  182. // It returns the leases along with the "as-of" etcd-index that can be used as the starting
  183. // point for etcd watch.
  184. func (m *EtcdManager) getLeases(ctx context.Context, network string) ([]Lease, uint64, error) {
  185. resp, err := m.registry.getSubnets(ctx, network)
  186. leases := []Lease{}
  187. index := uint64(0)
  188. switch {
  189. case err == nil:
  190. for _, node := range resp.Node.Nodes {
  191. sn, err := parseSubnetKey(node.Key)
  192. if err == nil {
  193. attrs := &LeaseAttrs{}
  194. if err = json.Unmarshal([]byte(node.Value), attrs); err == nil {
  195. exp := time.Time{}
  196. if node.Expiration != nil {
  197. exp = *node.Expiration
  198. }
  199. lease := Lease{
  200. Subnet: sn,
  201. Attrs: attrs,
  202. Expiration: exp,
  203. }
  204. leases = append(leases, lease)
  205. }
  206. }
  207. }
  208. index = resp.Node.ModifiedIndex
  209. case err.(*etcd.EtcdError).ErrorCode == etcdKeyNotFound:
  210. // key not found: treat it as empty set
  211. default:
  212. return nil, 0, err
  213. }
  214. return leases, index, nil
  215. }
  216. func (m *EtcdManager) RenewLease(ctx context.Context, network string, lease *Lease) error {
  217. attrBytes, err := json.Marshal(lease.Attrs)
  218. if err != nil {
  219. return err
  220. }
  221. // TODO(eyakubovich): propogate ctx into registry
  222. resp, err := m.registry.updateSubnet(ctx, network, lease.Key(), string(attrBytes), subnetTTL)
  223. if err != nil {
  224. return err
  225. }
  226. lease.Expiration = *resp.Node.Expiration
  227. return nil
  228. }
  229. func (m *EtcdManager) WatchLeases(ctx context.Context, network string, cursor interface{}) (WatchResult, error) {
  230. nextIndex := uint64(0)
  231. if cursor != nil {
  232. nextIndex = cursor.(uint64)
  233. }
  234. resp, err := m.registry.watchSubnets(ctx, network, nextIndex)
  235. switch {
  236. case err == nil:
  237. return parseSubnetWatchResponse(resp)
  238. case isIndexTooSmall(err):
  239. log.Warning("Watch of subnet leases failed because etcd index outside history window")
  240. return m.watchReset(ctx, network)
  241. default:
  242. return WatchResult{}, err
  243. }
  244. }
  245. func isIndexTooSmall(err error) bool {
  246. etcdErr, ok := err.(*etcd.EtcdError)
  247. return ok && etcdErr.ErrorCode == etcdEventIndexCleared
  248. }
  249. func parseSubnetWatchResponse(resp *etcd.Response) (WatchResult, error) {
  250. sn, err := parseSubnetKey(resp.Node.Key)
  251. if err != nil {
  252. return WatchResult{}, fmt.Errorf("error parsing subnet IP: %s", resp.Node.Key)
  253. }
  254. evt := Event{}
  255. switch resp.Action {
  256. case "delete", "expire":
  257. evt = Event{
  258. SubnetRemoved,
  259. Lease{Subnet: sn},
  260. }
  261. default:
  262. attrs := &LeaseAttrs{}
  263. err := json.Unmarshal([]byte(resp.Node.Value), attrs)
  264. if err != nil {
  265. return WatchResult{}, err
  266. }
  267. exp := time.Time{}
  268. if resp.Node.Expiration != nil {
  269. exp = *resp.Node.Expiration
  270. }
  271. evt = Event{
  272. SubnetAdded,
  273. Lease{
  274. Subnet: sn,
  275. Attrs: attrs,
  276. Expiration: exp,
  277. },
  278. }
  279. }
  280. cursor := resp.Node.ModifiedIndex + 1
  281. return WatchResult{
  282. Cursor: cursor,
  283. Events: []Event{evt},
  284. }, nil
  285. }
  286. // watchReset is called when incremental watch failed and we need to grab a snapshot
  287. func (m *EtcdManager) watchReset(ctx context.Context, network string) (WatchResult, error) {
  288. wr := WatchResult{}
  289. leases, index, err := m.getLeases(ctx, network)
  290. if err != nil {
  291. return wr, fmt.Errorf("failed to retrieve subnet leases: %v", err)
  292. }
  293. cursor := index + 1
  294. wr.Snapshot = leases
  295. wr.Cursor = cursor
  296. return wr, nil
  297. }
  298. func interrupted(cancel chan bool) bool {
  299. select {
  300. case <-cancel:
  301. return true
  302. default:
  303. return false
  304. }
  305. }