main.go 6.9 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211
  1. package main
  2. import (
  3. "bufio"
  4. "context"
  5. "encoding/base64"
  6. "encoding/json"
  7. "flag"
  8. "fmt"
  9. "io"
  10. "net/http"
  11. "os"
  12. "strings"
  13. "github.com/sirupsen/logrus"
  14. "github.com/containerd/containerd/log"
  15. "github.com/docker/docker/libnetwork"
  16. "github.com/docker/docker/libnetwork/diagnostic"
  17. "github.com/docker/docker/libnetwork/drivers/overlay"
  18. )
  19. const (
  20. readyPath = "http://%s:%d/ready"
  21. joinNetwork = "http://%s:%d/joinnetwork?nid=%s"
  22. leaveNetwork = "http://%s:%d/leavenetwork?nid=%s"
  23. clusterPeers = "http://%s:%d/clusterpeers?json"
  24. networkPeers = "http://%s:%d/networkpeers?nid=%s&json"
  25. dumpTable = "http://%s:%d/gettable?nid=%s&tname=%s&json"
  26. deleteEntry = "http://%s:%d/deleteentry?nid=%s&tname=%s&key=%s&json"
  27. )
  28. func httpIsOk(body io.ReadCloser) {
  29. b, err := io.ReadAll(body)
  30. if err != nil {
  31. log.G(context.TODO()).Fatalf("Failed the body parse %s", err)
  32. }
  33. if !strings.Contains(string(b), "OK") {
  34. log.G(context.TODO()).Fatalf("Server not ready %s", b)
  35. }
  36. body.Close()
  37. }
  38. func main() {
  39. ipPtr := flag.String("ip", "127.0.0.1", "ip address")
  40. portPtr := flag.Int("port", 2000, "port")
  41. networkPtr := flag.String("net", "", "target network")
  42. tablePtr := flag.String("t", "", "table to process <sd/overlay>")
  43. remediatePtr := flag.Bool("r", false, "perform remediation deleting orphan entries")
  44. joinPtr := flag.Bool("a", false, "join/leave network")
  45. verbosePtr := flag.Bool("v", false, "verbose output")
  46. flag.Parse()
  47. if *verbosePtr {
  48. logrus.SetLevel(logrus.DebugLevel)
  49. }
  50. if _, ok := os.LookupEnv("DIND_CLIENT"); !ok && *joinPtr {
  51. log.G(context.TODO()).Fatal("you are not using the client in docker in docker mode, the use of the -a flag can be disruptive, " +
  52. "please remove it (doc:https://github.com/docker/docker/libnetwork/blob/master/cmd/diagnostic/README.md)")
  53. }
  54. log.G(context.TODO()).Infof("Connecting to %s:%d checking ready", *ipPtr, *portPtr)
  55. resp, err := http.Get(fmt.Sprintf(readyPath, *ipPtr, *portPtr))
  56. if err != nil {
  57. log.G(context.TODO()).WithError(err).Fatalf("The connection failed")
  58. }
  59. httpIsOk(resp.Body)
  60. clusterPeers := fetchNodePeers(*ipPtr, *portPtr, "")
  61. var networkPeers map[string]string
  62. var joinedNetwork bool
  63. if *networkPtr != "" {
  64. if *joinPtr {
  65. log.G(context.TODO()).Infof("Joining the network:%q", *networkPtr)
  66. resp, err = http.Get(fmt.Sprintf(joinNetwork, *ipPtr, *portPtr, *networkPtr))
  67. if err != nil {
  68. log.G(context.TODO()).WithError(err).Fatalf("Failed joining the network")
  69. }
  70. httpIsOk(resp.Body)
  71. joinedNetwork = true
  72. }
  73. networkPeers = fetchNodePeers(*ipPtr, *portPtr, *networkPtr)
  74. if len(networkPeers) == 0 {
  75. log.G(context.TODO()).Warnf("There is no peer on network %q, check the network ID, and verify that is the non truncated version", *networkPtr)
  76. }
  77. }
  78. switch *tablePtr {
  79. case "sd":
  80. fetchTable(*ipPtr, *portPtr, *networkPtr, "endpoint_table", clusterPeers, networkPeers, *remediatePtr)
  81. case "overlay":
  82. fetchTable(*ipPtr, *portPtr, *networkPtr, "overlay_peer_table", clusterPeers, networkPeers, *remediatePtr)
  83. }
  84. if joinedNetwork {
  85. log.G(context.TODO()).Infof("Leaving the network:%q", *networkPtr)
  86. resp, err = http.Get(fmt.Sprintf(leaveNetwork, *ipPtr, *portPtr, *networkPtr))
  87. if err != nil {
  88. log.G(context.TODO()).WithError(err).Fatalf("Failed leaving the network")
  89. }
  90. httpIsOk(resp.Body)
  91. }
  92. }
  93. func fetchNodePeers(ip string, port int, network string) map[string]string {
  94. if network == "" {
  95. log.G(context.TODO()).Infof("Fetch cluster peers")
  96. } else {
  97. log.G(context.TODO()).Infof("Fetch peers network:%q", network)
  98. }
  99. var path string
  100. if network != "" {
  101. path = fmt.Sprintf(networkPeers, ip, port, network)
  102. } else {
  103. path = fmt.Sprintf(clusterPeers, ip, port)
  104. }
  105. resp, err := http.Get(path) //nolint:gosec // G107: Potential HTTP request made with variable url
  106. if err != nil {
  107. log.G(context.TODO()).WithError(err).Fatalf("Failed fetching path")
  108. }
  109. defer resp.Body.Close()
  110. body, err := io.ReadAll(resp.Body)
  111. if err != nil {
  112. log.G(context.TODO()).WithError(err).Fatalf("Failed the body parse")
  113. }
  114. output := diagnostic.HTTPResult{Details: &diagnostic.TablePeersResult{}}
  115. err = json.Unmarshal(body, &output)
  116. if err != nil {
  117. log.G(context.TODO()).WithError(err).Fatalf("Failed the json unmarshalling")
  118. }
  119. log.G(context.TODO()).Debugf("Parsing JSON response")
  120. result := make(map[string]string, output.Details.(*diagnostic.TablePeersResult).Length)
  121. for _, v := range output.Details.(*diagnostic.TablePeersResult).Elements {
  122. log.G(context.TODO()).Debugf("name:%s ip:%s", v.Name, v.IP)
  123. result[v.Name] = v.IP
  124. }
  125. return result
  126. }
  127. func fetchTable(ip string, port int, network, tableName string, clusterPeers, networkPeers map[string]string, remediate bool) {
  128. log.G(context.TODO()).Infof("Fetch %s table and check owners", tableName)
  129. resp, err := http.Get(fmt.Sprintf(dumpTable, ip, port, network, tableName))
  130. if err != nil {
  131. log.G(context.TODO()).WithError(err).Fatalf("Failed fetching endpoint table")
  132. }
  133. defer resp.Body.Close()
  134. body, err := io.ReadAll(resp.Body)
  135. if err != nil {
  136. log.G(context.TODO()).WithError(err).Fatalf("Failed the body parse")
  137. }
  138. output := diagnostic.HTTPResult{Details: &diagnostic.TableEndpointsResult{}}
  139. err = json.Unmarshal(body, &output)
  140. if err != nil {
  141. log.G(context.TODO()).WithError(err).Fatalf("Failed the json unmarshalling")
  142. }
  143. log.G(context.TODO()).Debug("Parsing data structures")
  144. var orphanKeys []string
  145. for _, v := range output.Details.(*diagnostic.TableEndpointsResult).Elements {
  146. decoded, err := base64.StdEncoding.DecodeString(v.Value)
  147. if err != nil {
  148. log.G(context.TODO()).WithError(err).Errorf("Failed decoding entry")
  149. continue
  150. }
  151. switch tableName {
  152. case "endpoint_table":
  153. var elem libnetwork.EndpointRecord
  154. elem.Unmarshal(decoded)
  155. log.G(context.TODO()).Debugf("key:%s value:%+v owner:%s", v.Key, elem, v.Owner)
  156. case "overlay_peer_table":
  157. var elem overlay.PeerRecord
  158. elem.Unmarshal(decoded)
  159. log.G(context.TODO()).Debugf("key:%s value:%+v owner:%s", v.Key, elem, v.Owner)
  160. }
  161. if _, ok := networkPeers[v.Owner]; !ok {
  162. log.G(context.TODO()).Warnf("The element with key:%s does not belong to any node on this network", v.Key)
  163. orphanKeys = append(orphanKeys, v.Key)
  164. }
  165. if _, ok := clusterPeers[v.Owner]; !ok {
  166. log.G(context.TODO()).Warnf("The element with key:%s does not belong to any node on this cluster", v.Key)
  167. }
  168. }
  169. if len(orphanKeys) > 0 && remediate {
  170. log.G(context.TODO()).Warnf("The following keys:%v results as orphan, do you want to proceed with the deletion (this operation is irreversible)? [Yes/No]", orphanKeys)
  171. reader := bufio.NewReader(os.Stdin)
  172. text, _ := reader.ReadString('\n')
  173. text = strings.ReplaceAll(text, "\n", "")
  174. if strings.Compare(text, "Yes") == 0 {
  175. for _, k := range orphanKeys {
  176. resp, err := http.Get(fmt.Sprintf(deleteEntry, ip, port, network, tableName, k))
  177. if err != nil {
  178. log.G(context.TODO()).WithError(err).Errorf("Failed deleting entry k:%s", k)
  179. break
  180. }
  181. resp.Body.Close()
  182. }
  183. } else {
  184. log.G(context.TODO()).Infof("Deletion skipped")
  185. }
  186. }
  187. }