haproxy-selfie/lua/prometheus.lua at master · AnchorFree/haproxy-selfie · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
--[[

 prom.translations table holds the mapping between the HAProxy internal names
 for metrics (as returned by Proxy.get_stats() and Server.get_stats()
 methods, see also
 https://github.com/haproxy/haproxy/blob/master/doc/management.txt,
 section "9.1 CSV Format") and the names we present to prometheus.

 Final string for each metric is built by the following formula:
 prom.prefix + variant + translation + optional labels + metric value.
 Allowed variants: "frontend", "backend", "server", "listener".
 See prom.build_metrics and prom.build_labels for additional details.
                                                                         ]]--

prom = {
  prefix = "haproxy",
  translations = {
    bin = "bytes_in_total",
    bout = "bytes_out_total",
    econ = "connection_errors_total",
    ereq = "request_errors_total",
    eresp = "response_errors_total",
    qcur = "current_queue",
    qmax = "max_queue",
    qlimit = "queue_limit",
    scur = "current_sessions",
    smax = "max_sessions",
    slim = "limit_sessions",
    stot = "total_sessions",
    dreq = "requests_denied_total",
    dresp = "responses_denied_total",
    wretr = "retry_warnings_total",
    wredis = "redispatch_warnings_total",
    weight = "weight",
    act = "active_servers",
    bck = "backup_servers",
    chkfail = "check_failures_total",
    chkdown = "check_transitions_total",
    lastchg = "last_transition_interval",
    throttle = "throttle_percent",
    downtime = "downtime_seconds_total",
    lbtot = "selected_times_total",
    rate = "current_session_rate",
    rate_lim = "limit_sessions",
    rate_max = "max_session_rate",
    ctime = "connect_time_average",
    rtime = "response_time_average",
    qtime = "queue_time_average",
    ttime = "session_time_average",
    status = { "up", function(metric)
                        local _, regex = Regex.new("^(UP|no check).*", true)
                        if regex:exec(metric) then
                            return 1
                        end
                        return 0
                     end },
    hrsp_1xx = { "http_responses_total", { code = "1xx" }},
    hrsp_2xx = { "http_responses_total", { code = "2xx" }},
    hrsp_3xx = { "http_responses_total", { code = "3xx" }},
    hrsp_4xx = { "http_responses_total", { code = "4xx" }},
    hrsp_5xx = { "http_responses_total", { code = "5xx" }},
    hrsp_other = { "http_responses_total", { code = "other" }},
    req_rate = "http_request_rate",
    req_rate_max = "http_max_request_rate",
    req_tot = "http_requests_total"
  },
  types = {
    counter = "^[CM]",
    gauge = "^[AaDGLRm]"
  }
}

-- build_labels converts a table with labels into a string.
prom.build_labels = function(labels)

    local l = core.concat()
    for k,v in pairs(labels) do
        l:add(k.."=\""..v.."\",")
    end
    local lstring = l:dump()
    return string.sub(lstring,1,-2)

end

-- get_type converts internal HAProxy metric type into
-- either "counter" or "gauge" for prometheus.
prom.get_type = function(mtype)

    if mtype:match(prom.types.counter) then
        return "counter"
    elseif mtype:match(prom.types.gauge) then
        return "gauge"
    end
    return "untyped"

end

-- build_metrics returns rendered prometheus metrics for a particular
-- proxy variant: -- listener, frontend, backend, server.
-- We intentionally don't include "#HELP ..." strings for metrics in the
-- output: prometheus does not need them, human beings can read the docs,
-- and it's always good to save ourselves some traffic.
prom.build_metrics = function(prom, variant)

    local response = core.concat()

    for n, t in pairs(prom.translations) do

        if stats.metrics[variant] and stats.metrics.has[variant][n] then

            local translation = ""
            local predefined_labels = {}
            local translation_func

            if type(t) == "table" then
                translation = t[1]
                if type(t[2]) == "function" then
                    translation_func = t[2]
                else
                    for k,v in pairs(t[2]) do
                        predefined_labels[k] = v
                    end
                end
            else
                translation = t
            end

            response:add("# TYPE "..prom.prefix.."_"..variant.."_"..translation.." "..prom.get_type(stats.metrics.has[variant][n]:sub(2,2)).."\n")
            for proxy_id, metrics in pairs(stats.metrics[variant]) do
                local labels = predefined_labels
                if variant == "server" then
                    labels.addr = metrics.addr
                    labels.server = metrics.svname
                    labels.backend = metrics.pxname
                else
                    labels[variant] = metrics.pxname
                end
                if metrics[n] then
                    local metric_value = metrics[n]
                    if translation_func then
                        metric_value = translation_func(metric_value)
                    end
                    response:add(prom.prefix.."_"..variant.."_"..translation.."{"..prom.build_labels(labels).."} "..metric_value.."\n")
                end
            end
        end
    end
    return response

end

return prom