Python Agent Meter Reporter

Important Note: Meter reporter is currently available to send in gRPC and Kafka protocol, HTTP protocol is not implemented yet (requires additional handler on SkyWalking OAP side).

Enabling the feature (default is enabled)

PVM Reporter is also by default enabled, meaning useful Python metrics such as thread count/GC info will be shown in OAP General Services - Instance - PVM Tab) If you really don’t need such a feature, disable them through config.agent_pvm_meter_reporter_active or SW_AGENT_PVM_METER_REPORTER_ACTIVE

config.agent_meter_reporter_active = True
# Or
os.environ['SW_AGENT_METER_REPORTER_ACTIVE'] = 'True'

or

export SW_AGENT_METER_REPORTER_ACTIVE=True

Disable the feature

os.environ['SW_AGENT_METER_REPORTER_ACTIVE'] = 'False'

or

export SW_AGENT_METER_REPORTER_ACTIVE=False

Counter

  • Counter API represents a single monotonically increasing counter, automatic collect data and report to backend.
builder = Counter.Builder('c2', CounterMode.INCREMENT, (("k1", "v1"), ("k2", "v2")))
# or this way
# builder = Counter.Builder('c2', CounterMode.INCREMENT).tag('key1', 'value1').tag('key2', 'value2')
c = builder.build()
c.increment(2)

Syntactic sugars

builder = Counter.Builder('c2', CounterMode.INCREMENT)
c = builder.build()

# increase Counter c by the time the with-wrapped codes consumed
with c.create_timer():
    # some codes may consume a certain time

builder = Counter.Builder('c3', CounterMode.INCREMENT)
c = builder.build()

# increase Counter c by num once counter_decorator_test gets called
@Counter.increase(name='c3', num=2)
def counter_decorator_test():
    # some codes

builder = Counter.Builder('c4', CounterMode.INCREMENT)
c = builder.build()

# increase Counter c by the time counter_decorator_test consumed
@Counter.timer(name='c4')
def counter_decorator_test(s):
    # some codes may consume a certain time
  1. Counter.Builder(name, tags) Create a new counter builder with the meter name and optional tags.
  2. Counter.tag(key: str, value) Mark a tag key/value pair.
  3. Counter.mode(mode: CounterMode) Change the counter mode, RATE mode means reporting rate to the backend.
  4. Counter.increment(count) Increment count to the Counter, It could be a positive value.

Gauge

  • Gauge API represents a single numerical value.
# producer: iterable object
builder = Gauge.Builder('g1', producer, (("key", "value")))
g = Builder.build()
  1. Gauge.Builder(name, tags) Create a new gauge builder with the meter name and iterable object, this iterable object need to produce numeric value.
  2. Gauge.tag(key: str, value) Mark a tag key/value pair.
  3. Gauge.build() Build a new Gauge which is collected and reported to the backend.

Histogram

  • Histogram API represents a summary sample observations with customize buckets.
builder = Histogram.Builder('h2', [i / 10 for i in range(10)], ("key", "value"))
h = builder.build()

Syntactic sugars

builder = Histogram.Builder('h3', [i / 10 for i in range(10)])
h = builder.build()

# Histogram h will record the time the with-wprapped codes consumed
with h.create_timer():
    # some codes may consume a certain time

builder = Histogram.Builder('h2', [i / 10 for i in range(10)])
h = builder.build()

# Histogram h will record the time histogram_decorator_test consumed
@Histogram.timer(name='h2')
def histogram_decorator_test(s):
    time.sleep(s)
  1. Histogram.Builder(name, tags) Create a new histogram builder with the meter name and optional tags.
  2. Histogram.tag(key: str, value) Mark a tag key/value pair.
  3. Histogram.minValue(value) Set up the minimal value of this histogram, default is 0.
  4. Histogram.build() Build a new Histogram which is collected and reported to the backend.
  5. Histogram.addValue(value) Add value into the histogram, automatically analyze what bucket count needs to be increment. rule: count into [step1, step2).