| 1 | // SPDX-License-Identifier: GPL-3.0-or-later |
| 2 | |
| 3 | package mongo |
| 4 | |
| 5 | import ( |
| 6 | "fmt" |
| 7 | "strings" |
| 8 | |
| 9 | "github.com/netdata/netdata/go/plugins/plugin/framework/collectorapi" |
| 10 | "github.com/netdata/netdata/go/plugins/plugin/go.d/pkg/oldmetrix" |
| 11 | ) |
| 12 | |
| 13 | // https://www.mongodb.com/docs/manual/reference/replica-states/#replica-set-member-states |
| 14 | var replicaSetMemberStates = map[string]int{ |
| 15 | "startup": 0, |
| 16 | "primary": 1, |
| 17 | "secondary": 2, |
| 18 | "recovering": 3, |
| 19 | "startup2": 5, |
| 20 | "unknown": 6, |
| 21 | "arbiter": 7, |
| 22 | "down": 8, |
| 23 | "rollback": 9, |
| 24 | "removed": 10, |
| 25 | } |
| 26 | |
| 27 | // TODO: deal with duplicates if we collect metrics from all cluster nodes |
| 28 | // should we only collect ReplSetStatus (at least by default) from primary nodes? (db.runCommand( { isMaster: 1 } )) |
| 29 | func (c *Collector) collectReplSetStatus(mx map[string]int64) error { |
| 30 | s, err := c.conn.replSetGetStatus() |
| 31 | if err != nil { |
| 32 | return fmt.Errorf("error get status of the replica set from mongo: %s", err) |
| 33 | } |
| 34 | |
| 35 | seen := make(map[string]documentReplSetMember) |
| 36 | |
| 37 | for _, member := range s.Members { |
| 38 | seen[member.Name] = member |
| 39 | |
| 40 | px := fmt.Sprintf("repl_set_member_%s_", member.Name) |
| 41 | |
| 42 | mx[px+"replication_lag"] = s.Date.Sub(member.OptimeDate).Milliseconds() |
| 43 | |
| 44 | for k, v := range replicaSetMemberStates { |
| 45 | mx[px+"state_"+k] = oldmetrix.Bool(member.State == v) |
| 46 | } |
| 47 | |
| 48 | mx[px+"health_status_up"] = oldmetrix.Bool(member.Health == 1) |
| 49 | mx[px+"health_status_down"] = oldmetrix.Bool(member.Health == 0) |
| 50 | |
| 51 | if member.Self == nil { |
| 52 | mx[px+"uptime"] = member.Uptime |
| 53 | if v := member.LastHeartbeatRecv; v != nil && !v.IsZero() { |
| 54 | mx[px+"heartbeat_latency"] = s.Date.Sub(*v).Milliseconds() |
| 55 | } |
| 56 | if v := member.PingMs; v != nil { |
| 57 | mx[px+"ping_rtt"] = *v |
| 58 | } |
| 59 | } |
| 60 | } |
| 61 | |
| 62 | for name, member := range seen { |
| 63 | if !c.replSetMembers[name] { |
| 64 | c.replSetMembers[name] = true |
| 65 | c.Debugf("new replica set member '%s': adding charts", name) |
| 66 | c.addReplSetMemberCharts(member) |
| 67 | } |
| 68 | } |
| 69 | |
| 70 | for name := range c.replSetMembers { |
| 71 | if _, ok := seen[name]; !ok { |
| 72 | delete(c.replSetMembers, name) |
| 73 | c.Debugf("stale replica set member '%s': removing charts", name) |
| 74 | c.removeReplSetMemberCharts(name) |
| 75 | } |
| 76 | } |
| 77 | |
| 78 | return nil |
| 79 | } |
| 80 | |
| 81 | func (c *Collector) addReplSetMemberCharts(v documentReplSetMember) { |
| 82 | charts := chartsTmplReplSetMember.Copy() |
| 83 | |
| 84 | if v.Self != nil { |
| 85 | _ = charts.Remove(chartTmplReplSetMemberHeartbeatLatencyTime.ID) |
| 86 | _ = charts.Remove(chartTmplReplSetMemberPingRTTTime.ID) |
| 87 | _ = charts.Remove(chartTmplReplSetMemberUptime.ID) |
| 88 | } |
| 89 | |
| 90 | for _, chart := range *charts { |
| 91 | chart.ID = fmt.Sprintf(chart.ID, v.Name) |
| 92 | chart.Labels = []collectorapi.Label{ |
| 93 | {Key: "repl_set_member", Value: v.Name}, |
| 94 | } |
| 95 | for _, dim := range chart.Dims { |
| 96 | dim.ID = fmt.Sprintf(dim.ID, v.Name) |
| 97 | } |
| 98 | } |
| 99 | |
| 100 | if err := c.Charts().Add(*charts...); err != nil { |
| 101 | c.Warning(err) |
| 102 | } |
| 103 | } |
| 104 | |
| 105 | func (c *Collector) removeReplSetMemberCharts(name string) { |
| 106 | px := fmt.Sprintf("%s%s_", chartPxReplSetMember, name) |
| 107 | |
| 108 | for _, chart := range *c.Charts() { |
| 109 | if strings.HasPrefix(chart.ID, px) { |
| 110 | chart.MarkRemove() |
| 111 | chart.MarkNotCreated() |
| 112 | } |
| 113 | } |
| 114 | } |