时间点指标告诉您现在发生了什么。时间序列数据告诉您发生了什么变化、何时发生变化以及某些内容是否有出现问题的趋势。随着时间的推移，跟踪验证码解决率、延迟和成本，以便在影响管道之前捕获性能下降情况。
追踪什么
| 公制 | 类型 | 为什么 |
|---|---|---|
| 解决率(%) | 测量 | 检测提供商质量变化 |
| 求解延迟（毫秒） | 直方图 | 发现减速、计划超时 |
| 按代码的错误率 | 柜台 | 识别新出现的错误模式 |
| 每次解决的成本（美元） | 测量 | 预算跟踪、异常检测 |
| 队列深度 | 测量 | 容量规划 |
| 令牌在使用前已过期 | 柜台 | TTL调谐信号 |
| API余额 | 测量 | 重新填充触发器 |
Prometheus + Python（推送网关）
仪器你的解算器
import os
import time
import requests
from prometheus_client import CollectorRegistry, Counter, Histogram, Gauge, push_to_gateway
registry = CollectorRegistry()
SOLVE_TOTAL = Counter(
"captcha_solve_total", "Total CAPTCHA solve attempts",
["type", "status"], registry=registry
)
SOLVE_LATENCY = Histogram(
"captcha_solve_latency_seconds", "CAPTCHA solve latency",
["type"], buckets=[5, 10, 15, 20, 30, 45, 60, 90, 120],
registry=registry
)
SOLVE_COST = Counter(
"captcha_solve_cost_dollars", "Total cost of CAPTCHA solves",
["type"], registry=registry
)
API_BALANCE = Gauge(
"captcha_api_balance_dollars", "CaptchaAI account balance",
registry=registry
)
API_KEY = os.environ["CAPTCHAAI_API_KEY"]
PUSHGATEWAY = os.environ.get("PUSHGATEWAY_URL", "localhost:9091")
def solve_with_metrics(sitekey, pageurl, captcha_type="recaptcha_v2"):
start = time.time()
resp = requests.post("https://ocr.captchaai.com/in.php", data={
"key": API_KEY,
"method": "userrecaptcha",
"googlekey": sitekey,
"pageurl": pageurl,
"json": 1
})
data = resp.json()
if data.get("status") != 1:
SOLVE_TOTAL.labels(type=captcha_type, status="submit_error").inc()
push_metrics()
return {"error": data.get("request")}
captcha_id = data["request"]
for _ in range(60):
time.sleep(5)
result = requests.get("https://ocr.captchaai.com/res.php", params={
"key": API_KEY, "action": "get",
"id": captcha_id, "json": 1
}).json()
if result.get("status") == 1:
elapsed = time.time() - start
SOLVE_TOTAL.labels(type=captcha_type, status="solved").inc()
SOLVE_LATENCY.labels(type=captcha_type).observe(elapsed)
SOLVE_COST.labels(type=captcha_type).inc(0.00299)
push_metrics()
return {"solution": result["request"]}
if result.get("request") != "CAPCHA_NOT_READY":
SOLVE_TOTAL.labels(type=captcha_type, status="error").inc()
push_metrics()
return {"error": result.get("request")}
SOLVE_TOTAL.labels(type=captcha_type, status="timeout").inc()
push_metrics()
return {"error": "TIMEOUT"}
def push_metrics():
try:
push_to_gateway(PUSHGATEWAY, job="captcha_solver", registry=registry)
except Exception:
pass # Don't fail solving because metrics push failed
def update_balance():
resp = requests.get("https://ocr.captchaai.com/res.php", params={
"key": API_KEY, "action": "getbalance"
})
try:
balance = float(resp.text)
API_BALANCE.set(balance)
push_metrics()
except ValueError:
pass
普罗米修斯查询
# Success rate over last hour
rate(captcha_solve_total{status="solved"}[1h])
/ rate(captcha_solve_total[1h]) * 100
# P95 solve latency
histogram_quantile(0.95, rate(captcha_solve_latency_seconds_bucket[1h]))
# Error rate by type
rate(captcha_solve_total{status="error"}[1h])
# Hourly cost
increase(captcha_solve_cost_dollars_total[1h])
InfluxDB + Python
写入求解指标
from influxdb_client import InfluxDBClient, Point
from influxdb_client.client.write_api import SYNCHRONOUS
INFLUX_URL = os.environ.get("INFLUX_URL", "http://localhost:8086")
INFLUX_TOKEN = os.environ.get("INFLUX_TOKEN", "")
INFLUX_ORG = os.environ.get("INFLUX_ORG", "captcha")
INFLUX_BUCKET = os.environ.get("INFLUX_BUCKET", "captcha_metrics")
influx_client = InfluxDBClient(url=INFLUX_URL, token=INFLUX_TOKEN, org=INFLUX_ORG)
write_api = influx_client.write_api(write_options=SYNCHRONOUS)
def record_solve_metric(captcha_type, status, elapsed_ms, cost=0.0, error=None):
point = (
Point("captcha_solve")
.tag("type", captcha_type)
.tag("status", status)
.field("elapsed_ms", elapsed_ms)
.field("cost", cost)
.field("success", 1 if status == "solved" else 0)
)
if error:
point = point.tag("error_code", error)
write_api.write(bucket=INFLUX_BUCKET, record=point)
def record_balance(balance):
point = Point("captcha_balance").field("balance", balance)
write_api.write(bucket=INFLUX_BUCKET, record=point)
InfluxDB 查询 (Flux)
// Success rate over last 24 hours (1-hour windows)
from(bucket: "captcha_metrics")
|> range(start: -24h)
|> filter(fn: (r) => r._measurement == "captcha_solve" and r._field == "success")
|> aggregateWindow(every: 1h, fn: mean)
|> map(fn: (r) => ({r with _value: r._value * 100.0}))
|> yield(name: "success_rate")
// Average solve time by type
from(bucket: "captcha_metrics")
|> range(start: -24h)
|> filter(fn: (r) => r._measurement == "captcha_solve" and r._field == "elapsed_ms" and r.status == "solved")
|> group(columns: ["type"])
|> aggregateWindow(every: 1h, fn: mean)
|> yield(name: "avg_latency")
// Cumulative cost
from(bucket: "captcha_metrics")
|> range(start: -24h)
|> filter(fn: (r) => r._measurement == "captcha_solve" and r._field == "cost")
|> cumulativeSum()
|> yield(name: "cumulative_cost")
JavaScript 实现（普罗米修斯）
const client = require("prom-client");
const axios = require("axios");
const register = new client.Registry();
const API_KEY = process.env.CAPTCHAAI_API_KEY;
const solveTotal = new client.Counter({
name: "captcha_solve_total",
help: "Total CAPTCHA solve attempts",
labelNames: ["type", "status"],
registers: [register],
});
const solveLatency = new client.Histogram({
name: "captcha_solve_latency_seconds",
help: "CAPTCHA solve latency",
labelNames: ["type"],
buckets: [5, 10, 15, 20, 30, 45, 60, 90, 120],
registers: [register],
});
async function solveWithMetrics(sitekey, pageurl, type = "recaptcha_v2") {
const start = Date.now();
const submit = await axios.post("https://ocr.captchaai.com/in.php", null, {
params: { key: API_KEY, method: "userrecaptcha", googlekey: sitekey, pageurl, json: 1 },
});
if (submit.data.status !== 1) {
solveTotal.inc({ type, status: "submit_error" });
return { error: submit.data.request };
}
const captchaId = submit.data.request;
for (let i = 0; i < 60; i++) {
await new Promise((r) => setTimeout(r, 5000));
const poll = await axios.get("https://ocr.captchaai.com/res.php", {
params: { key: API_KEY, action: "get", id: captchaId, json: 1 },
});
if (poll.data.status === 1) {
const elapsed = (Date.now() - start) / 1000;
solveTotal.inc({ type, status: "solved" });
solveLatency.observe({ type }, elapsed);
return { solution: poll.data.request };
}
if (poll.data.request !== "CAPCHA_NOT_READY") {
solveTotal.inc({ type, status: "error" });
return { error: poll.data.request };
}
}
solveTotal.inc({ type, status: "timeout" });
return { error: "TIMEOUT" };
}
// Expose metrics endpoint
const express = require("express");
const app = express();
app.get("/metrics", async (req, res) => {
res.set("Content-Type", register.contentType);
res.end(await register.metrics());
});
app.listen(9090);
数据库比较
| 特征 | 普罗米修斯 | InfluxDB | 时标数据库 |
|---|---|---|---|
| 最适合 | 运营监控 | 物联网/高基数指标 | 基于 SQL 的分析 |
| 查询语言 | 普罗姆QL | 通量 | SQL |
| 保留 | 基于配置 | 基于政策 | PostgreSQL 保留 |
| Grafana 集成 | 本国的 | 本国的 | 本国的 |
| 学习曲线 | 低的 | 中等的 | 低（如果你懂 SQL） |
| 自托管 | 是的 | 是的 | 是（PostgreSQL 扩展） |
故障排除
| 问题 | 原因 | 处理方式 |
|---|---|---|
| 仪表板中的指标差距 | 推送网关收不到数据 | 检查求解器和推送网关之间的网络 |
| 延迟直方图显示错误的百分位数 | 存储桶边界与工作负载不匹配 | 调整桶：[5, 10, 15, 20, 30, 45, 60, 90, 120] 用于验证码解决 |
| 成本指标与实际支出不符 | 每个验证码类型的价格不同 | 按类型划分的标签成本；使用实际的每种类型定价 |
| 基数过多 | 标签值太多 | 将标签限制为 type、status、error_code |
常问问题
我应该使用哪个时间序列数据库？
如果您已经使用 Prometheus 进行基础设施监控，只需添加 CAPTCHA 指标即可。如果您想要独立的指标存储，则可以使用 InfluxDB。如果您想要对时间序列数据进行 SQL 查询，则可以使用 TimescaleDB。
我应该保留 CAPTCHA 指标多久？
将高分辨率数据（每秒）保留 7 天，汇总数据（每小时）保留 90 天，并无限期保留每日摘要。这可以平衡存储成本与趋势可见性。
我可以自动检测解决率下降吗？
是的。设置滚动平均值警报 - 例如，当 1 小时成功率低于 90% 或 P95 延迟超过 45 秒时发出警报。Prometheus Alertmanager 和 InfluxDB 警报都支持此功能。
下一步
随着时间的推移跟踪您的验证码解决性能 –”获取您的 CaptchaAI API 密钥并开始收集指标。
相关指南：
