main.go 6.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208
  1. package main
  2. import (
  3. "bufio"
  4. "encoding/base64"
  5. "encoding/json"
  6. "flag"
  7. "fmt"
  8. "io"
  9. "net/http"
  10. "os"
  11. "strings"
  12. "github.com/docker/docker/libnetwork"
  13. "github.com/docker/docker/libnetwork/diagnostic"
  14. "github.com/docker/docker/libnetwork/drivers/overlay"
  15. "github.com/sirupsen/logrus"
  16. )
  17. const (
  18. readyPath = "http://%s:%d/ready"
  19. joinNetwork = "http://%s:%d/joinnetwork?nid=%s"
  20. leaveNetwork = "http://%s:%d/leavenetwork?nid=%s"
  21. clusterPeers = "http://%s:%d/clusterpeers?json"
  22. networkPeers = "http://%s:%d/networkpeers?nid=%s&json"
  23. dumpTable = "http://%s:%d/gettable?nid=%s&tname=%s&json"
  24. deleteEntry = "http://%s:%d/deleteentry?nid=%s&tname=%s&key=%s&json"
  25. )
  26. func httpIsOk(body io.ReadCloser) {
  27. b, err := io.ReadAll(body)
  28. if err != nil {
  29. logrus.Fatalf("Failed the body parse %s", err)
  30. }
  31. if !strings.Contains(string(b), "OK") {
  32. logrus.Fatalf("Server not ready %s", b)
  33. }
  34. body.Close()
  35. }
  36. func main() {
  37. ipPtr := flag.String("ip", "127.0.0.1", "ip address")
  38. portPtr := flag.Int("port", 2000, "port")
  39. networkPtr := flag.String("net", "", "target network")
  40. tablePtr := flag.String("t", "", "table to process <sd/overlay>")
  41. remediatePtr := flag.Bool("r", false, "perform remediation deleting orphan entries")
  42. joinPtr := flag.Bool("a", false, "join/leave network")
  43. verbosePtr := flag.Bool("v", false, "verbose output")
  44. flag.Parse()
  45. if *verbosePtr {
  46. logrus.SetLevel(logrus.DebugLevel)
  47. }
  48. if _, ok := os.LookupEnv("DIND_CLIENT"); !ok && *joinPtr {
  49. logrus.Fatal("you are not using the client in docker in docker mode, the use of the -a flag can be disruptive, " +
  50. "please remove it (doc:https://github.com/docker/docker/libnetwork/blob/master/cmd/diagnostic/README.md)")
  51. }
  52. logrus.Infof("Connecting to %s:%d checking ready", *ipPtr, *portPtr)
  53. resp, err := http.Get(fmt.Sprintf(readyPath, *ipPtr, *portPtr))
  54. if err != nil {
  55. logrus.WithError(err).Fatalf("The connection failed")
  56. }
  57. httpIsOk(resp.Body)
  58. clusterPeers := fetchNodePeers(*ipPtr, *portPtr, "")
  59. var networkPeers map[string]string
  60. var joinedNetwork bool
  61. if *networkPtr != "" {
  62. if *joinPtr {
  63. logrus.Infof("Joining the network:%q", *networkPtr)
  64. resp, err = http.Get(fmt.Sprintf(joinNetwork, *ipPtr, *portPtr, *networkPtr))
  65. if err != nil {
  66. logrus.WithError(err).Fatalf("Failed joining the network")
  67. }
  68. httpIsOk(resp.Body)
  69. joinedNetwork = true
  70. }
  71. networkPeers = fetchNodePeers(*ipPtr, *portPtr, *networkPtr)
  72. if len(networkPeers) == 0 {
  73. logrus.Warnf("There is no peer on network %q, check the network ID, and verify that is the non truncated version", *networkPtr)
  74. }
  75. }
  76. switch *tablePtr {
  77. case "sd":
  78. fetchTable(*ipPtr, *portPtr, *networkPtr, "endpoint_table", clusterPeers, networkPeers, *remediatePtr)
  79. case "overlay":
  80. fetchTable(*ipPtr, *portPtr, *networkPtr, "overlay_peer_table", clusterPeers, networkPeers, *remediatePtr)
  81. }
  82. if joinedNetwork {
  83. logrus.Infof("Leaving the network:%q", *networkPtr)
  84. resp, err = http.Get(fmt.Sprintf(leaveNetwork, *ipPtr, *portPtr, *networkPtr))
  85. if err != nil {
  86. logrus.WithError(err).Fatalf("Failed leaving the network")
  87. }
  88. httpIsOk(resp.Body)
  89. }
  90. }
  91. func fetchNodePeers(ip string, port int, network string) map[string]string {
  92. if network == "" {
  93. logrus.Infof("Fetch cluster peers")
  94. } else {
  95. logrus.Infof("Fetch peers network:%q", network)
  96. }
  97. var path string
  98. if network != "" {
  99. path = fmt.Sprintf(networkPeers, ip, port, network)
  100. } else {
  101. path = fmt.Sprintf(clusterPeers, ip, port)
  102. }
  103. resp, err := http.Get(path) //nolint:gosec // G107: Potential HTTP request made with variable url
  104. if err != nil {
  105. logrus.WithError(err).Fatalf("Failed fetching path")
  106. }
  107. defer resp.Body.Close()
  108. body, err := io.ReadAll(resp.Body)
  109. if err != nil {
  110. logrus.WithError(err).Fatalf("Failed the body parse")
  111. }
  112. output := diagnostic.HTTPResult{Details: &diagnostic.TablePeersResult{}}
  113. err = json.Unmarshal(body, &output)
  114. if err != nil {
  115. logrus.WithError(err).Fatalf("Failed the json unmarshalling")
  116. }
  117. logrus.Debugf("Parsing JSON response")
  118. result := make(map[string]string, output.Details.(*diagnostic.TablePeersResult).Length)
  119. for _, v := range output.Details.(*diagnostic.TablePeersResult).Elements {
  120. logrus.Debugf("name:%s ip:%s", v.Name, v.IP)
  121. result[v.Name] = v.IP
  122. }
  123. return result
  124. }
  125. func fetchTable(ip string, port int, network, tableName string, clusterPeers, networkPeers map[string]string, remediate bool) {
  126. logrus.Infof("Fetch %s table and check owners", tableName)
  127. resp, err := http.Get(fmt.Sprintf(dumpTable, ip, port, network, tableName))
  128. if err != nil {
  129. logrus.WithError(err).Fatalf("Failed fetching endpoint table")
  130. }
  131. defer resp.Body.Close()
  132. body, err := io.ReadAll(resp.Body)
  133. if err != nil {
  134. logrus.WithError(err).Fatalf("Failed the body parse")
  135. }
  136. output := diagnostic.HTTPResult{Details: &diagnostic.TableEndpointsResult{}}
  137. err = json.Unmarshal(body, &output)
  138. if err != nil {
  139. logrus.WithError(err).Fatalf("Failed the json unmarshalling")
  140. }
  141. logrus.Debug("Parsing data structures")
  142. var orphanKeys []string
  143. for _, v := range output.Details.(*diagnostic.TableEndpointsResult).Elements {
  144. decoded, err := base64.StdEncoding.DecodeString(v.Value)
  145. if err != nil {
  146. logrus.WithError(err).Errorf("Failed decoding entry")
  147. continue
  148. }
  149. switch tableName {
  150. case "endpoint_table":
  151. var elem libnetwork.EndpointRecord
  152. elem.Unmarshal(decoded)
  153. logrus.Debugf("key:%s value:%+v owner:%s", v.Key, elem, v.Owner)
  154. case "overlay_peer_table":
  155. var elem overlay.PeerRecord
  156. elem.Unmarshal(decoded)
  157. logrus.Debugf("key:%s value:%+v owner:%s", v.Key, elem, v.Owner)
  158. }
  159. if _, ok := networkPeers[v.Owner]; !ok {
  160. logrus.Warnf("The element with key:%s does not belong to any node on this network", v.Key)
  161. orphanKeys = append(orphanKeys, v.Key)
  162. }
  163. if _, ok := clusterPeers[v.Owner]; !ok {
  164. logrus.Warnf("The element with key:%s does not belong to any node on this cluster", v.Key)
  165. }
  166. }
  167. if len(orphanKeys) > 0 && remediate {
  168. logrus.Warnf("The following keys:%v results as orphan, do you want to proceed with the deletion (this operation is irreversible)? [Yes/No]", orphanKeys)
  169. reader := bufio.NewReader(os.Stdin)
  170. text, _ := reader.ReadString('\n')
  171. text = strings.Replace(text, "\n", "", -1)
  172. if strings.Compare(text, "Yes") == 0 {
  173. for _, k := range orphanKeys {
  174. resp, err := http.Get(fmt.Sprintf(deleteEntry, ip, port, network, tableName, k))
  175. if err != nil {
  176. logrus.WithError(err).Errorf("Failed deleting entry k:%s", k)
  177. break
  178. }
  179. resp.Body.Close()
  180. }
  181. } else {
  182. logrus.Infof("Deletion skipped")
  183. }
  184. }
  185. }