# Copyright (C) 2018 The Software Heritage developers
# See the AUTHORS file at the top-level directory of this distribution
# License: GNU General Public License version 3, or any later version
# See top-level LICENSE file for more information
# Initially imported from https://github.com/DataDog/datadogpy/
# at revision 62b3a3e89988dc18d78c282fe3ff5d1813917436
#
# Copyright (c) 2015, Datadog <info@datadoghq.com>
# All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
# * Redistributions of source code must retain the above copyright
# notice, this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright
# notice, this list of conditions and the following disclaimer in the
# documentation and/or other materials provided with the distribution.
# * Neither the name of Datadog nor the names of its contributors may be
# used to endorse or promote products derived from this software without
# specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL <COPYRIGHT HOLDER> BE LIABLE FOR ANY
# DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
# (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
# LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
# ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#
#
# Vastly adapted for integration in swh.core:
#
# - Removed python < 3.5 compat code
# - trimmed the imports down to be a single module
# - adjust some options:
# - drop unix socket connection option
# - add environment variable support for setting the statsd host and
# port (pulled the idea from the main python statsd module)
# - only send timer metrics in milliseconds (that's what
# prometheus-statsd-exporter expects)
# - drop DataDog-specific metric types (that are unsupported in
# prometheus-statsd-exporter)
# - made the tags a dict instead of a list (prometheus-statsd-exporter only
# supports tags with a value, mirroring prometheus)
# - switch from time.time to time.monotonic
# - improve unit test coverage
# - documentation cleanup
from asyncio import iscoroutinefunction
from contextlib import contextmanager
from functools import wraps
import itertools
import logging
import os
from random import random
import re
import socket
import threading
from time import monotonic
from typing import Collection, Dict, Optional
import warnings
log = logging.getLogger("swh.core.statsd")
[docs]
class TimedContextManagerDecorator(object):
"""
A context manager and a decorator which will report the elapsed time in
the context OR in a function call.
Attributes:
elapsed (float): the elapsed time at the point of completion
"""
def __init__(
self, statsd, metric=None, error_metric=None, tags=None, sample_rate=1
):
self.statsd = statsd
self.metric = metric
self.error_metric = error_metric
self.tags = tags or {}
self.sample_rate = sample_rate
self.elapsed = None # this is for testing purpose
def __call__(self, func):
"""
Decorator which returns the elapsed time of the function call.
Default to the function name if metric was not provided.
"""
if not self.metric:
self.metric = "%s.%s" % (func.__module__, func.__name__)
# Coroutines
if iscoroutinefunction(func):
@wraps(func)
async def wrapped_co(*args, **kwargs):
start = monotonic()
try:
result = await func(*args, **kwargs)
except BaseException as e:
self._send_error(error_type=type(e).__name__)
raise
self._send(start)
return result
return wrapped_co
# Others
@wraps(func)
def wrapped(*args, **kwargs):
start = monotonic()
try:
result = func(*args, **kwargs)
except BaseException as e:
self._send_error(error_type=type(e).__name__)
raise
self._send(start)
return result
return wrapped
def __enter__(self):
if not self.metric:
raise TypeError("Cannot used timed without a metric!")
self._start = monotonic()
return self
def __exit__(self, type, value, traceback):
# Report the elapsed time of the context manager if no error.
if type is None:
self._send(self._start)
else:
self._send_error(error_type=type.__name__)
def _send(self, start):
elapsed = (monotonic() - start) * 1000
self.statsd.timing(
self.metric, elapsed, tags=self.tags, sample_rate=self.sample_rate
)
self.elapsed = elapsed
def _send_error(self, error_type=None):
if self.error_metric is None:
self.error_metric = self.metric + "_error_count"
if error_type is not None:
tags = {**self.tags, "error_type": error_type}
else:
tags = self.tags
self.statsd.increment(self.error_metric, tags=tags)
[docs]
def start(self):
"""Start the timer"""
self.__enter__()
[docs]
def stop(self):
"""Stop the timer, send the metric value"""
self.__exit__(None, None, None)
[docs]
class Statsd(object):
"""Initialize a client to send metrics to a StatsD server.
Arguments:
host (str): the host of the StatsD server. Defaults to localhost.
port (int): the port of the StatsD server. Defaults to 8125.
max_buffer_size (int): Maximum number of metrics to buffer before
sending to the server if sending metrics in batch
namespace (str): Namespace to prefix all metric names
constant_tags (Dict[str, str]): Tags to attach to all metrics
Note:
This class also supports the following environment variables:
STATSD_HOST
Override the default host of the statsd server
STATSD_PORT
Override the default port of the statsd server
STATSD_TAGS
Tags to attach to every metric reported. Example value:
"label:value,other_label:other_value"
"""
def __init__(
self,
*,
host=None,
port=None,
max_buffer_size=50,
namespace=None,
constant_tags=None,
):
# Connection
if host is None:
host = os.environ.get("STATSD_HOST") or "localhost"
self.host = host
if port is None:
port = os.environ.get("STATSD_PORT") or 8125
self.port = int(port)
# Socket
self._socket = None
self.lock = threading.Lock()
self.max_buffer_size = max_buffer_size
self._send = self._send_to_server
self.encoding = "utf-8"
# Tags
self.constant_tags = {}
tags_envvar = os.environ.get("STATSD_TAGS", "")
for tag in tags_envvar.split(","):
if not tag:
continue
if ":" not in tag:
warnings.warn(
f"STATSD_TAGS needs to be in 'key:value' format, not {tag!r}",
UserWarning,
)
continue
k, v = tag.split(":", 1)
# look for a possible env var substitution, using $NAME or ${NAME} format
m = re.match(r"^[$]([{])?(?P<envvar>\w+)(?(1)[}]|)$", v)
if m:
envvar = m.group("envvar")
if envvar in os.environ:
v = os.environ[envvar]
self.constant_tags[k] = v
if constant_tags:
self.constant_tags.update(
{str(k): str(v) for k, v in constant_tags.items()}
)
# Namespace
if namespace is not None:
namespace = str(namespace)
self.namespace = namespace
def __enter__(self):
self.open_buffer(self.max_buffer_size)
return self
def __exit__(self, type, value, traceback):
self.close_buffer()
[docs]
def gauge(self, metric, value, tags=None, sample_rate=1):
"""
Record the value of a gauge, optionally setting a list of tags and a
sample rate.
>>> statsd.gauge('users.online', 123)
>>> statsd.gauge('active.connections', 1001, tags={"protocol": "http"})
"""
return self._report(metric, "g", value, tags, sample_rate)
[docs]
def increment(self, metric, value=1, tags=None, sample_rate=1):
"""
Increment a counter, optionally setting a value, tags and a sample
rate.
>>> statsd.increment('page.views')
>>> statsd.increment('files.transferred', 124)
"""
self._report(metric, "c", value, tags, sample_rate)
[docs]
def decrement(self, metric, value=1, tags=None, sample_rate=1):
"""
Decrement a counter, optionally setting a value, tags and a sample
rate.
>>> statsd.decrement('files.remaining')
>>> statsd.decrement('active.connections', 2)
"""
metric_value = -value if value else value
self._report(metric, "c", metric_value, tags, sample_rate)
[docs]
def histogram(self, metric, value, tags=None, sample_rate=1):
"""
Sample a histogram value, optionally setting tags and a sample rate.
>>> statsd.histogram('uploaded.file.size', 1445)
>>> statsd.histogram('file.count', 26, tags={"filetype": "python"})
"""
self._report(metric, "h", value, tags, sample_rate)
[docs]
def timing(self, metric, value, tags=None, sample_rate=1):
"""
Record a timing, optionally setting tags and a sample rate.
>>> statsd.timing("query.response.time", 1234)
"""
self._report(metric, "ms", value, tags, sample_rate)
[docs]
def timed(self, metric=None, error_metric=None, tags=None, sample_rate=1):
"""
A decorator or context manager that will measure the distribution of a
function's/context's run time. Optionally specify a list of tags or a
sample rate. If the metric is not defined as a decorator, the module
name and function name will be used. The metric is required as a
context manager.
::
@statsd.timed('user.query.time', sample_rate=0.5)
def get_user(user_id):
# Do what you need to ...
pass
# Is equivalent to ...
with statsd.timed('user.query.time', sample_rate=0.5):
# Do what you need to ...
pass
# Is equivalent to ...
start = time.monotonic()
try:
get_user(user_id)
finally:
statsd.timing('user.query.time', time.monotonic() - start)
"""
return TimedContextManagerDecorator(
statsd=self,
metric=metric,
error_metric=error_metric,
tags=tags,
sample_rate=sample_rate,
)
[docs]
def set(self, metric, value, tags=None, sample_rate=1):
"""
Sample a set value.
>>> statsd.set('visitors.uniques', 999)
"""
self._report(metric, "s", value, tags, sample_rate)
@property
def socket(self):
"""
Return a connected socket.
Note: connect the socket before assigning it to the class instance to
avoid bad thread race conditions.
"""
with self.lock:
if not self._socket:
sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
sock.connect((self.host, self.port))
self._socket = sock
return self._socket
[docs]
def open_buffer(self, max_buffer_size=50):
"""
Open a buffer to send a batch of metrics in one packet.
You can also use this as a context manager.
>>> with Statsd() as batch:
... batch.gauge('users.online', 123)
... batch.gauge('active.connections', 1001)
"""
self.max_buffer_size = max_buffer_size
self.buffer = []
self._send = self._send_to_buffer
[docs]
def close_buffer(self):
"""
Flush the buffer and switch back to single metric packets.
"""
self._send = self._send_to_server
if self.buffer:
# Only send packets if there are packets to send
self._flush_buffer()
[docs]
def close_socket(self):
"""
Closes connected socket if connected.
"""
with self.lock:
if self._socket:
self._socket.close()
self._socket = None
def _report(self, metric, metric_type, value, tags, sample_rate):
"""
Create a metric packet and send it.
"""
if value is None:
return
if sample_rate != 1 and random() > sample_rate:
return
# Resolve the full tag list
tags = self._add_constant_tags(tags)
# Create/format the metric packet
payload = "%s%s:%s|%s%s%s" % (
(self.namespace + ".") if self.namespace else "",
metric,
value,
metric_type,
("|@" + str(sample_rate)) if sample_rate != 1 else "",
(
("|#" + ",".join("%s:%s" % (k, v) for (k, v) in sorted(tags.items())))
if tags
else ""
),
)
# Send it
self._send(payload)
def _send_to_server(self, packet):
try:
# If set, use socket directly
self.socket.send(packet.encode("utf-8"))
except socket.timeout:
return
except socket.error:
log.debug(
"Error submitting statsd packet."
" Dropping the packet and closing the socket."
)
self.close_socket()
def _send_to_buffer(self, packet):
self.buffer.append(packet)
if len(self.buffer) >= self.max_buffer_size:
self._flush_buffer()
def _flush_buffer(self):
self._send_to_server("\n".join(self.buffer))
self.buffer = []
def _add_constant_tags(self, tags):
return {
str(k): str(v)
for k, v in itertools.chain(
self.constant_tags.items(),
(tags if tags else {}).items(),
)
}
[docs]
@contextmanager
def status_gauge(
self,
metric_name: str,
statuses: Collection[str],
tags: Optional[Dict[str, str]] = None,
):
"""Context manager to keep track of status changes as a gauge
In addition to the `metric_name` and `tags` arguments, it expects a
list of `statuses` to declare which statuses are possible, and returns
a callable as context manager. This callable takes ones of the possible
statuses as argument. Typical usage would be:
>>> with statsd.status_gauge(
"metric_name", ["starting", "processing", "waiting"]) as set_status:
set_status("starting")
# ...
set_status("waiting")
# ...
"""
if tags is None:
tags = {}
current_status: Optional[str] = None
# reset status gauges to make sure they do not "leak"
for status in statuses:
self.gauge(metric_name, 0, {**tags, "status": status})
def set_status(new_status: str):
nonlocal current_status
assert isinstance(tags, dict)
if new_status not in statuses:
raise ValueError(f"{new_status} not in {statuses}")
if current_status and new_status != current_status:
self.gauge(metric_name, 0, {**tags, "status": current_status})
current_status = new_status
self.gauge(metric_name, 1, {**tags, "status": current_status})
try:
yield set_status
finally:
# reset gauges on exit
for status in statuses:
self.gauge(metric_name, 0, {**tags, "status": status})
statsd = Statsd()
