|
| 1 | +package main |
| 2 | + |
| 3 | +import ( |
| 4 | + "bufio" |
| 5 | + "encoding/base64" |
| 6 | + "encoding/json" |
| 7 | + "flag" |
| 8 | + "fmt" |
| 9 | + "io" |
| 10 | + "io/ioutil" |
| 11 | + "net/http" |
| 12 | + "os" |
| 13 | + "strings" |
| 14 | + |
| 15 | + "github.com/docker/libnetwork" |
| 16 | + "github.com/docker/libnetwork/diagnose" |
| 17 | + "github.com/docker/libnetwork/drivers/overlay" |
| 18 | + "github.com/sirupsen/logrus" |
| 19 | +) |
| 20 | + |
| 21 | +const ( |
| 22 | + readyPath = "http://%s:%d/ready" |
| 23 | + joinNetwork = "http://%s:%d/joinnetwork?nid=%s" |
| 24 | + leaveNetwork = "http://%s:%d/leavenetwork?nid=%s" |
| 25 | + clusterPeers = "http://%s:%d/clusterpeers?json" |
| 26 | + networkPeers = "http://%s:%d/networkpeers?nid=%s&json" |
| 27 | + dumpTable = "http://%s:%d/gettable?nid=%s&tname=%s&json" |
| 28 | + deleteEntry = "http://%s:%d/deleteentry?nid=%s&tname=%s&key=%s&json" |
| 29 | +) |
| 30 | + |
| 31 | +func httpIsOk(body io.ReadCloser) { |
| 32 | + b, err := ioutil.ReadAll(body) |
| 33 | + if err != nil { |
| 34 | + logrus.Fatalf("Failed the body parse %s", err) |
| 35 | + } |
| 36 | + if !strings.Contains(string(b), "OK") { |
| 37 | + logrus.Fatalf("Server not ready %s", b) |
| 38 | + } |
| 39 | + body.Close() |
| 40 | +} |
| 41 | + |
| 42 | +func main() { |
| 43 | + ipPtr := flag.String("ip", "127.0.0.1", "ip address") |
| 44 | + portPtr := flag.Int("port", 2000, "port") |
| 45 | + networkPtr := flag.String("net", "", "target network") |
| 46 | + tablePtr := flag.String("t", "", "table to process <sd/overlay>") |
| 47 | + remediatePtr := flag.Bool("r", false, "perform remediation deleting orphan entries") |
| 48 | + verbosePtr := flag.Bool("v", false, "verbose output") |
| 49 | + |
| 50 | + flag.Parse() |
| 51 | + |
| 52 | + if *verbosePtr { |
| 53 | + logrus.SetLevel(logrus.DebugLevel) |
| 54 | + } |
| 55 | + |
| 56 | + logrus.Infof("Connecting to %s:%d checking ready", *ipPtr, *portPtr) |
| 57 | + resp, err := http.Get(fmt.Sprintf(readyPath, *ipPtr, *portPtr)) |
| 58 | + if err != nil { |
| 59 | + logrus.WithError(err).Fatalf("The connection failed") |
| 60 | + } |
| 61 | + httpIsOk(resp.Body) |
| 62 | + |
| 63 | + clusterPeers := fetchNodePeers(*ipPtr, *portPtr, "") |
| 64 | + var networkPeers map[string]string |
| 65 | + var joinedNetwork bool |
| 66 | + if *networkPtr != "" { |
| 67 | + logrus.Infof("Joining the network:%s", *networkPtr) |
| 68 | + resp, err = http.Get(fmt.Sprintf(joinNetwork, *ipPtr, *portPtr, *networkPtr)) |
| 69 | + if err != nil { |
| 70 | + logrus.WithError(err).Fatalf("Failed joining the network") |
| 71 | + } |
| 72 | + httpIsOk(resp.Body) |
| 73 | + networkPeers = fetchNodePeers(*ipPtr, *portPtr, *networkPtr) |
| 74 | + joinedNetwork = true |
| 75 | + } |
| 76 | + |
| 77 | + switch *tablePtr { |
| 78 | + case "sd": |
| 79 | + fetchTable(*ipPtr, *portPtr, *networkPtr, "endpoint_table", clusterPeers, networkPeers, *remediatePtr) |
| 80 | + case "overlay": |
| 81 | + fetchTable(*ipPtr, *portPtr, *networkPtr, "overlay_peer_table", clusterPeers, networkPeers, *remediatePtr) |
| 82 | + } |
| 83 | + |
| 84 | + if joinedNetwork { |
| 85 | + resp, err = http.Get(fmt.Sprintf(leaveNetwork, *ipPtr, *portPtr, *networkPtr)) |
| 86 | + if err != nil { |
| 87 | + logrus.WithError(err).Fatalf("Failed leaving the network") |
| 88 | + } |
| 89 | + httpIsOk(resp.Body) |
| 90 | + } |
| 91 | +} |
| 92 | + |
| 93 | +func fetchNodePeers(ip string, port int, network string) map[string]string { |
| 94 | + logrus.Infof("Fetch peers %s", network) |
| 95 | + var path string |
| 96 | + if network != "" { |
| 97 | + path = fmt.Sprintf(networkPeers, ip, port, network) |
| 98 | + } else { |
| 99 | + path = fmt.Sprintf(clusterPeers, ip, port) |
| 100 | + } |
| 101 | + |
| 102 | + resp, err := http.Get(path) |
| 103 | + if err != nil { |
| 104 | + logrus.WithError(err).Fatalf("Failed fetching path") |
| 105 | + } |
| 106 | + defer resp.Body.Close() |
| 107 | + body, err := ioutil.ReadAll(resp.Body) |
| 108 | + if err != nil { |
| 109 | + logrus.WithError(err).Fatalf("Failed the body parse") |
| 110 | + } |
| 111 | + |
| 112 | + output := diagnose.HTTPResult{Details: &diagnose.TablePeersResult{}} |
| 113 | + err = json.Unmarshal(body, &output) |
| 114 | + if err != nil { |
| 115 | + logrus.WithError(err).Fatalf("Failed the json unmarshalling") |
| 116 | + } |
| 117 | + |
| 118 | + logrus.Debugf("Parsing JSON response") |
| 119 | + result := make(map[string]string, output.Details.(*diagnose.TablePeersResult).Length) |
| 120 | + for _, v := range output.Details.(*diagnose.TablePeersResult).Elements { |
| 121 | + logrus.Debugf("name:%s ip:%s", v.Name, v.IP) |
| 122 | + result[v.Name] = v.IP |
| 123 | + } |
| 124 | + return result |
| 125 | +} |
| 126 | + |
| 127 | +func fetchTable(ip string, port int, network, tableName string, networkPeers, clusterPeers map[string]string, remediate bool) { |
| 128 | + logrus.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 | + logrus.WithError(err).Fatalf("Failed fetching endpoint table") |
| 132 | + } |
| 133 | + defer resp.Body.Close() |
| 134 | + body, err := ioutil.ReadAll(resp.Body) |
| 135 | + if err != nil { |
| 136 | + logrus.WithError(err).Fatalf("Failed the body parse") |
| 137 | + } |
| 138 | + |
| 139 | + output := diagnose.HTTPResult{Details: &diagnose.TableEndpointsResult{}} |
| 140 | + err = json.Unmarshal(body, &output) |
| 141 | + if err != nil { |
| 142 | + logrus.WithError(err).Fatalf("Failed the json unmarshalling") |
| 143 | + } |
| 144 | + |
| 145 | + logrus.Debug("Parsing data structures") |
| 146 | + var orphanKeys []string |
| 147 | + for _, v := range output.Details.(*diagnose.TableEndpointsResult).Elements { |
| 148 | + decoded, err := base64.StdEncoding.DecodeString(v.Value) |
| 149 | + if err != nil { |
| 150 | + logrus.WithError(err).Errorf("Failed decoding entry") |
| 151 | + continue |
| 152 | + } |
| 153 | + switch tableName { |
| 154 | + case "endpoint_table": |
| 155 | + var elem libnetwork.EndpointRecord |
| 156 | + elem.Unmarshal(decoded) |
| 157 | + logrus.Debugf("key:%s value:%+v owner:%s", v.Key, elem, v.Owner) |
| 158 | + case "overlay_peer_table": |
| 159 | + var elem overlay.PeerRecord |
| 160 | + elem.Unmarshal(decoded) |
| 161 | + logrus.Debugf("key:%s value:%+v owner:%s", v.Key, elem, v.Owner) |
| 162 | + } |
| 163 | + |
| 164 | + if _, ok := networkPeers[v.Owner]; !ok { |
| 165 | + logrus.Warnf("The element with key:%s does not belong to any node on this network", v.Key) |
| 166 | + orphanKeys = append(orphanKeys, v.Key) |
| 167 | + } |
| 168 | + if _, ok := clusterPeers[v.Owner]; !ok { |
| 169 | + logrus.Warnf("The element with key:%s does not belong to any node on this cluster", v.Key) |
| 170 | + } |
| 171 | + } |
| 172 | + |
| 173 | + if len(orphanKeys) > 0 && remediate { |
| 174 | + logrus.Warnf("The following keys:%v results as orphan, do you want to proceed with the deletion (this operation is irreversible)? [Yes/No]", orphanKeys) |
| 175 | + reader := bufio.NewReader(os.Stdin) |
| 176 | + text, _ := reader.ReadString('\n') |
| 177 | + text = strings.Replace(text, "\n", "", -1) |
| 178 | + if strings.Compare(text, "Yes") == 0 { |
| 179 | + for _, k := range orphanKeys { |
| 180 | + resp, err := http.Get(fmt.Sprintf(deleteEntry, ip, port, network, tableName, k)) |
| 181 | + if err != nil { |
| 182 | + logrus.WithError(err).Errorf("Failed deleting entry k:%s", k) |
| 183 | + break |
| 184 | + } |
| 185 | + resp.Body.Close() |
| 186 | + } |
| 187 | + } else { |
| 188 | + logrus.Infof("Deletion skipped") |
| 189 | + } |
| 190 | + } |
| 191 | +} |
0 commit comments