| 1 | // SPDX-License-Identifier: GPL-3.0-or-later |
| 2 | |
| 3 | package consul |
| 4 | |
| 5 | import ( |
| 6 | "net/http" |
| 7 | "time" |
| 8 | |
| 9 | "github.com/netdata/netdata/go/plugins/plugin/go.d/pkg/oldmetrix" |
| 10 | ) |
| 11 | |
| 12 | const ( |
| 13 | // https://developer.hashicorp.com/consul/api-docs/operator/autopilot#read-health |
| 14 | urlPathOperationAutopilotHealth = "/v1/operator/autopilot/health" |
| 15 | ) |
| 16 | |
| 17 | type autopilotHealth struct { |
| 18 | Servers []struct { |
| 19 | ID string |
| 20 | SerfStatus string |
| 21 | Leader bool |
| 22 | LastContact string |
| 23 | Healthy bool |
| 24 | Voter bool |
| 25 | StableSince time.Time |
| 26 | } |
| 27 | } |
| 28 | |
| 29 | func (c *Collector) collectAutopilotHealth(mx map[string]int64) error { |
| 30 | req, err := c.createRequest(urlPathOperationAutopilotHealth) |
| 31 | if err != nil { |
| 32 | return err |
| 33 | } |
| 34 | |
| 35 | var health autopilotHealth |
| 36 | |
| 37 | // The HTTP status code will indicate the health of the cluster: 200 is healthy, 429 is unhealthy. |
| 38 | // https://github.com/hashicorp/consul/blob/c7ef04c5979dbc311ff3c67b7bf3028a93e8b0f1/agent/operator_endpoint.go#L325 |
| 39 | if err := c.client(http.StatusTooManyRequests).RequestJSON(req, &health); err != nil { |
| 40 | return err |
| 41 | } |
| 42 | |
| 43 | for _, srv := range health.Servers { |
| 44 | if srv.ID == c.cfg.Config.NodeID { |
| 45 | // SerfStatus: alive, left, failed or none: |
| 46 | // https://github.com/hashicorp/consul/blob/c7ef04c5979dbc311ff3c67b7bf3028a93e8b0f1/agent/consul/operator_autopilot_endpoint.go#L124-L133 |
| 47 | mx["autopilot_server_sefStatus_alive"] = oldmetrix.Bool(srv.SerfStatus == "alive") |
| 48 | mx["autopilot_server_sefStatus_left"] = oldmetrix.Bool(srv.SerfStatus == "left") |
| 49 | mx["autopilot_server_sefStatus_failed"] = oldmetrix.Bool(srv.SerfStatus == "failed") |
| 50 | mx["autopilot_server_sefStatus_none"] = oldmetrix.Bool(srv.SerfStatus == "none") |
| 51 | // https://github.com/hashicorp/raft-autopilot/blob/d936f51c374c3b7902d5e4fdafe9f7d8d199ea53/types.go#L110 |
| 52 | mx["autopilot_server_healthy_yes"] = oldmetrix.Bool(srv.Healthy) |
| 53 | mx["autopilot_server_healthy_no"] = oldmetrix.Bool(!srv.Healthy) |
| 54 | mx["autopilot_server_voter_yes"] = oldmetrix.Bool(srv.Voter) |
| 55 | mx["autopilot_server_voter_no"] = oldmetrix.Bool(!srv.Voter) |
| 56 | mx["autopilot_server_stable_time"] = int64(time.Since(srv.StableSince).Seconds()) |
| 57 | mx["autopilot_server_stable_time"] = int64(time.Since(srv.StableSince).Seconds()) |
| 58 | if !srv.Leader { |
| 59 | if v, err := time.ParseDuration(srv.LastContact); err == nil { |
| 60 | mx["autopilot_server_lastContact_leader"] = v.Milliseconds() |
| 61 | } |
| 62 | } |
| 63 | |
| 64 | break |
| 65 | } |
| 66 | } |
| 67 | |
| 68 | return nil |
| 69 | } |