File size: 8,562 Bytes
e3278e4 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 |
# used for monitoring litellm services health on `/metrics` endpoint on LiteLLM Proxy
#### What this does ####
# On success + failure, log events to Prometheus for litellm / adjacent services (litellm, redis, postgres, llm api providers)
from typing import List, Optional, Union
from litellm._logging import print_verbose, verbose_logger
from litellm.types.integrations.prometheus import LATENCY_BUCKETS
from litellm.types.services import ServiceLoggerPayload, ServiceTypes
FAILED_REQUESTS_LABELS = ["error_class", "function_name"]
class PrometheusServicesLogger:
# Class variables or attributes
litellm_service_latency = None # Class-level attribute to store the Histogram
def __init__(
self,
mock_testing: bool = False,
**kwargs,
):
try:
try:
from prometheus_client import REGISTRY, Counter, Histogram
except ImportError:
raise Exception(
"Missing prometheus_client. Run `pip install prometheus-client`"
)
self.Histogram = Histogram
self.Counter = Counter
self.REGISTRY = REGISTRY
verbose_logger.debug("in init prometheus services metrics")
self.services = [item.value for item in ServiceTypes]
self.payload_to_prometheus_map = (
{}
) # store the prometheus histogram/counter we need to call for each field in payload
for service in self.services:
histogram = self.create_histogram(service, type_of_request="latency")
counter_failed_request = self.create_counter(
service,
type_of_request="failed_requests",
additional_labels=FAILED_REQUESTS_LABELS,
)
counter_total_requests = self.create_counter(
service, type_of_request="total_requests"
)
self.payload_to_prometheus_map[service] = [
histogram,
counter_failed_request,
counter_total_requests,
]
self.prometheus_to_amount_map: dict = (
{}
) # the field / value in ServiceLoggerPayload the object needs to be incremented by
### MOCK TESTING ###
self.mock_testing = mock_testing
self.mock_testing_success_calls = 0
self.mock_testing_failure_calls = 0
except Exception as e:
print_verbose(f"Got exception on init prometheus client {str(e)}")
raise e
def is_metric_registered(self, metric_name) -> bool:
for metric in self.REGISTRY.collect():
if metric_name == metric.name:
return True
return False
def _get_metric(self, metric_name):
"""
Helper function to get a metric from the registry by name.
"""
return self.REGISTRY._names_to_collectors.get(metric_name)
def create_histogram(self, service: str, type_of_request: str):
metric_name = "litellm_{}_{}".format(service, type_of_request)
is_registered = self.is_metric_registered(metric_name)
if is_registered:
return self._get_metric(metric_name)
return self.Histogram(
metric_name,
"Latency for {} service".format(service),
labelnames=[service],
buckets=LATENCY_BUCKETS,
)
def create_counter(
self,
service: str,
type_of_request: str,
additional_labels: Optional[List[str]] = None,
):
metric_name = "litellm_{}_{}".format(service, type_of_request)
is_registered = self.is_metric_registered(metric_name)
if is_registered:
return self._get_metric(metric_name)
return self.Counter(
metric_name,
"Total {} for {} service".format(type_of_request, service),
labelnames=[service] + (additional_labels or []),
)
def observe_histogram(
self,
histogram,
labels: str,
amount: float,
):
assert isinstance(histogram, self.Histogram)
histogram.labels(labels).observe(amount)
def increment_counter(
self,
counter,
labels: str,
amount: float,
additional_labels: Optional[List[str]] = [],
):
assert isinstance(counter, self.Counter)
if additional_labels:
counter.labels(labels, *additional_labels).inc(amount)
else:
counter.labels(labels).inc(amount)
def service_success_hook(self, payload: ServiceLoggerPayload):
if self.mock_testing:
self.mock_testing_success_calls += 1
if payload.service.value in self.payload_to_prometheus_map:
prom_objects = self.payload_to_prometheus_map[payload.service.value]
for obj in prom_objects:
if isinstance(obj, self.Histogram):
self.observe_histogram(
histogram=obj,
labels=payload.service.value,
amount=payload.duration,
)
elif isinstance(obj, self.Counter) and "total_requests" in obj._name:
self.increment_counter(
counter=obj,
labels=payload.service.value,
amount=1, # LOG TOTAL REQUESTS TO PROMETHEUS
)
def service_failure_hook(self, payload: ServiceLoggerPayload):
if self.mock_testing:
self.mock_testing_failure_calls += 1
if payload.service.value in self.payload_to_prometheus_map:
prom_objects = self.payload_to_prometheus_map[payload.service.value]
for obj in prom_objects:
if isinstance(obj, self.Counter):
self.increment_counter(
counter=obj,
labels=payload.service.value,
amount=1, # LOG ERROR COUNT / TOTAL REQUESTS TO PROMETHEUS
)
async def async_service_success_hook(self, payload: ServiceLoggerPayload):
"""
Log successful call to prometheus
"""
if self.mock_testing:
self.mock_testing_success_calls += 1
if payload.service.value in self.payload_to_prometheus_map:
prom_objects = self.payload_to_prometheus_map[payload.service.value]
for obj in prom_objects:
if isinstance(obj, self.Histogram):
self.observe_histogram(
histogram=obj,
labels=payload.service.value,
amount=payload.duration,
)
elif isinstance(obj, self.Counter) and "total_requests" in obj._name:
self.increment_counter(
counter=obj,
labels=payload.service.value,
amount=1, # LOG TOTAL REQUESTS TO PROMETHEUS
)
async def async_service_failure_hook(
self,
payload: ServiceLoggerPayload,
error: Union[str, Exception],
):
if self.mock_testing:
self.mock_testing_failure_calls += 1
error_class = error.__class__.__name__
function_name = payload.call_type
if payload.service.value in self.payload_to_prometheus_map:
prom_objects = self.payload_to_prometheus_map[payload.service.value]
for obj in prom_objects:
# increment both failed and total requests
if isinstance(obj, self.Counter):
if "failed_requests" in obj._name:
self.increment_counter(
counter=obj,
labels=payload.service.value,
# log additional_labels=["error_class", "function_name"], used for debugging what's going wrong with the DB
additional_labels=[error_class, function_name],
amount=1, # LOG ERROR COUNT TO PROMETHEUS
)
else:
self.increment_counter(
counter=obj,
labels=payload.service.value,
amount=1, # LOG TOTAL REQUESTS TO PROMETHEUS
)
|