File size: 7,187 Bytes
f3b0a91 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 | """Series builders: integer counters and gauges with diurnal traffic and jitter (U9).
All counters start at large random offsets so Prometheus' counter zero-clamp never triggers, and all
series in a scenario share sample timestamps, so extrapolation factors cancel in ratios.
"""
from __future__ import annotations
import math
import random
from fractions import Fraction as F
from .promsim import STALE, Series
LES = ["0.05", "0.1", "0.25", "0.5", "1", "2.5", "+Inf"]
GRPC_ERR = ["Unavailable", "Internal", "DeadlineExceeded", "Unknown", "DataLoss"]
# PromQL selectors and reference predicates per SLI family. `bad` selects error series; latency
# SLIs count requests slower than the SLO threshold as bad (1 - fast/total).
FAMILIES = {
"http": {
"metric": "http_requests_total",
"bad_sel": 'code=~"5.."',
"bad": lambda l: l.get("code", "").startswith("5"),
},
"grpc": {
"metric": "grpc_server_handled_total",
"bad_sel": 'grpc_code=~"' + "|".join(GRPC_ERR) + '"',
"bad": lambda l: l.get("grpc_code") in GRPC_ERR,
},
"latency": {
"metric": "http_request_duration_seconds_bucket",
"count": "http_request_duration_seconds_count",
},
}
def diurnal(rng: random.Random, minutes: int) -> list[float]:
amp, phase = rng.uniform(0.08, 0.3), rng.uniform(0, 1440)
return [1 + amp * math.sin(2 * math.pi * (m + phase) / 1440) for m in range(minutes + 1)]
def totals(rng: random.Random, base: int, minutes: int, mult: list[float] | None = None) -> list[int]:
"""Per-minute request counts; index 0 is the scrape at t=0 and carries no increment."""
mult = mult or diurnal(rng, minutes)
return [0] + [max(1, int(round(base * mult[m] + rng.randint(-2, 2)))) for m in range(1, minutes + 1)]
def bads(rng: random.Random, tot: list[int], ratios: list[F | float]) -> list[int]:
"""Integer bad counts per minute following ratio[m], with +-1 jitter (never above the total)."""
out = [0]
carry = 0.0
for m in range(1, len(tot)):
exact = float(ratios[m]) * tot[m] + carry
b = int(math.floor(exact))
carry = exact - b
if 0 < b < tot[m] and float(ratios[m]) * tot[m] > 4:
b += rng.choice((-1, 0, 0, 1))
out.append(min(max(b, 0), tot[m]))
return out
def counter(offset: int, incs: list[int]) -> list[int]:
vals, cur = [], offset
for i, inc in enumerate(incs):
cur += inc if i else 0
vals.append(cur)
return vals
def split(rng: random.Random, incs: list[int], k: int) -> list[list[int]]:
"""Split per-minute increments across k pods with fixed random weights (integers, exact sums)."""
if k == 1:
return [list(incs)]
w = [rng.uniform(0.6, 1.4) for _ in range(k)]
s = sum(w)
parts = [[0] * len(incs) for _ in range(k)]
for m, inc in enumerate(incs):
acc = 0
for i in range(k - 1):
p = int(inc * w[i] / s)
parts[i][m] = p
acc += p
parts[k - 1][m] = inc - acc
return parts
def pod_labels(rng: random.Random, service: str, i: int, ns: str) -> dict[str, str]:
h = "".join(rng.choice("bcdfghjklmnpqrstvwxz2456789") for _ in range(5))
return {"service": service, "namespace": ns, "pod": f"{service}-{h}",
"instance": f"10.{rng.randint(0, 255)}.{rng.randint(0, 255)}.{i + 2}:8080"}
def sli_series(rng: random.Random, family: str, service: str, tot: list[int], bad: list[int],
pods: int, ns: str, slo_le: str = "0.5") -> list[Series]:
out: list[Series] = []
tparts, bparts = split(rng, tot, pods), split(rng, bad, pods)
for i in range(pods):
base = pod_labels(rng, service, i, ns)
good = [t - b for t, b in zip(tparts[i], bparts[i])]
if family == "http":
noise = [g // 25 for g in good] # 4xx traffic counts toward total, not errors
ok = [g - n for g, n in zip(good, noise)]
e1 = [b - b // 3 for b in bparts[i]]
e2 = [b // 3 for b in bparts[i]]
for code, incs in (("200", ok), ("404", noise), ("500", e1), ("503", e2)):
out.append(Series("http_requests_total", {**base, "code": code, "method": "GET"},
counter(rng.randint(10**5, 10**6) if code == "200" else rng.randint(500, 5000), incs)))
elif family == "grpc":
noise = [g // 30 for g in good]
ok = [g - n for g, n in zip(good, noise)]
e1 = [b - b // 4 for b in bparts[i]]
e2 = [b // 4 for b in bparts[i]]
for code, incs in (("OK", ok), ("NotFound", noise), ("Unavailable", e1), ("Internal", e2)):
out.append(Series("grpc_server_handled_total", {**base, "grpc_code": code,
"grpc_method": "Get"}, counter(rng.randint(500, 10**6), incs)))
else:
out.extend(histogram(rng, base, good, bparts[i], slo_le))
return out
def histogram(rng: random.Random, base: dict, fast: list[int], slow: list[int], thr_le: str) -> list[Series]:
"""Bucket counters: `fast` requests land in buckets <= thr_le, `slow` ones above it."""
idx = LES.index(thr_le)
lo, hi = LES[: idx + 1], LES[idx + 1:]
wl = [rng.uniform(0.5, 1.5) for _ in lo]
wh = [rng.uniform(0.5, 1.5) for _ in hi]
per_bucket = {le: [0] * len(fast) for le in LES}
for m in range(len(fast)):
for group, weights, n in ((lo, wl, fast[m]), (hi, wh, slow[m])):
s, acc = sum(weights), 0
for j, le in enumerate(group):
c = n - acc if j == len(group) - 1 else int(n * weights[j] / s)
per_bucket[le][m] += c
acc += c
out, cum = [], [0] * len(fast)
offset = rng.randint(10**4, 10**5)
for le in LES:
cum = [a + b for a, b in zip(cum, per_bucket[le])]
out.append(Series("http_request_duration_seconds_bucket", {**base, "le": le},
counter(offset, cum)))
offset += rng.randint(0, 50)
count_offset = out[-1].values[0]
out.append(Series("http_request_duration_seconds_count", dict(base), counter(count_offset, cum)))
return out
def gauge(values: list[int], labels: dict, metric: str) -> Series:
return Series(metric, dict(labels), list(values))
def up_series(rng: random.Random, service: str, instances: int, ns: str,
down_from: dict[int, tuple[int, int]] | None = None, minutes: int = 30,
stale_from: int | None = None) -> list[Series]:
"""`up` per instance; down_from maps instance index -> (start, end) minutes at 0."""
out = []
for i in range(instances):
lab = pod_labels(rng, service, i, ns)
lab = {"service": service, "instance": lab["instance"], "job": service, "namespace": ns}
vals: list = [1] * (minutes + 1)
if down_from and i in down_from:
a, b = down_from[i]
for m in range(a, min(b, minutes + 1)):
vals[m] = 0
if stale_from is not None:
vals = vals[:stale_from] + [STALE] + [None] * (minutes - stale_from)
out.append(Series("up", lab, vals))
return out
|