def test_tracing_load():
"""
Test that all of the traces sent through the agent get the proper service
correlation datapoint.
"""
port = random.randint(5001, 20000)
with Agent.run(
dedent(
f"""
hostname: "testhost"
writer:
sendTraceHostCorrelationMetrics: true
traceHostCorrelationMetricsInterval: 1s
staleServiceTimeout: 7s
monitors:
- type: trace-forwarder
listenAddress: localhost:{port}
"""
)
) as agent:
assert wait_for(p(tcp_port_open_locally, port)), "trace forwarder port never opened!"
for i in range(0, 100):
spans = _test_trace()
spans[0]["localEndpoint"]["serviceName"] += f"-{i}"
spans[1]["localEndpoint"]["serviceName"] += f"-{i}"
resp = retry_on_ebadf(
lambda: requests.post(
f"http://localhost:{port}/v1/trace",
headers={"Content-Type": "application/json"},
data=json.dumps(spans), # pylint:disable=cell-var-from-loop
)
)()
assert resp.status_code == 200
for i in range(0, 100):
assert wait_for(
p(
has_datapoint,
agent.fake_services,
metric_name="sf.int.service.heartbeat",
dimensions={"sf_hasService": f"myapp-{i}", "host": "testhost"},
)
), "Didn't get host correlation datapoint"
assert wait_for(
p(
has_datapoint,
agent.fake_services,
metric_name="sf.int.service.heartbeat",
dimensions={"sf_hasService": f"file-server-{i}", "host": "testhost"},
)
), "Didn't get host correlation datapoint"
time.sleep(10)
agent.fake_services.reset_datapoints()
assert ensure_never(
p(has_datapoint, agent.fake_services, metric_name="sf.int.service.heartbeat"), timeout_seconds=5
), "Got infra correlation metric when it should have been expired"
def test_filesystems_fstype_filter():
expected_metrics = [
"df_complex.free",
"df_complex.used",
"percent_bytes.free",
"percent_bytes.used",
"disk.utilization",
]
if sys.platform == "linux":
expected_metrics.extend(["df_inodes.free", "df_inodes.used", "percent_inodes.free", "percent_inodes.used"])
with Agent.run(
"""
procPath: /proc
monitors:
- type: filesystems
fsTypes:
- "!*"
"""
) as agent:
assert wait_for(
p(has_any_metric_or_dim, agent.fake_services, ["disk.summary_utilization"], []), timeout_seconds=60
), "timed out waiting for metrics and/or dimensions!"
assert ensure_never(lambda: has_any_metric_or_dim(agent.fake_services, expected_metrics, []))
assert not has_log_message(agent.output.lower(), "error"), "error found in agent output!"
def test_tracing_output():
"""
Test that the basic trace writer and service tracker work
"""
port = random.randint(5001, 20000)
with Agent.run(
dedent(f"""
hostname: "testhost"
writer:
traceHostCorrelationMetricsInterval: 1s
staleServiceTimeout: 5s
monitors:
- type: trace-forwarder
listenAddress: localhost:{port}
""")) as agent:
assert wait_for(p(tcp_port_open_locally,
port)), "trace forwarder port never opened!"
resp = requests.post(
f"http://localhost:{port}/v1/trace",
headers={"Content-Type": "application/json"},
data=json.dumps(_test_trace()),
)
assert resp.status_code == 200
assert wait_for(
p(has_trace_span, agent.fake_services,
tags={"env": "prod"})), "Didn't get span tag"
assert wait_for(p(has_trace_span, agent.fake_services,
name="fetch")), "Didn't get span name"
assert wait_for(
p(
has_datapoint,
agent.fake_services,
metric_name="sf.int.service.heartbeat",
dimensions={
"sf_hasService": "myapp",
"host": "testhost"
},
)), "Didn't get host correlation datapoint"
# Service names expire after 5s in the config provided in this test
time.sleep(8)
agent.fake_services.reset_datapoints()
assert ensure_never(
p(has_datapoint,
agent.fake_services,
metric_name="sf.int.service.heartbeat"),
timeout_seconds=5
), "Got infra correlation metric when it should have been expired"
def test_netio_filter():
forbidden_metrics = METADATA.included_metrics - {"network.total"}
with Agent.run(
"""
procPath: /proc
monitors:
- type: net-io
interfaces:
- "!*"
"""
) as agent:
assert wait_for(
p(has_datapoint, agent.fake_services, metric_name="network.total"), timeout_seconds=60
), "timed out waiting for metrics and/or dimensions!"
assert ensure_never(p(any_metric_found, agent.fake_services, forbidden_metrics))
assert not has_log_message(agent.output.lower(), "error"), "error found in agent output!"
def test_netio_filter():
expected_metrics = get_monitor_metrics_from_selfdescribe("net-io")
try:
expected_metrics.remove("network.total")
except KeyError:
pass
with run_agent("""
procPath: /proc
monitors:
- type: net-io
interfaces:
- "!*"
""") as [backend, get_output, _]:
assert wait_for(p(has_any_metric_or_dim, backend, ["network.total"],
[]),
timeout_seconds=60
), "timed out waiting for metrics and/or dimensions!"
assert ensure_never(
lambda: has_any_metric_or_dim(backend, expected_metrics, []))
assert not has_log_message(get_output().lower(),
"error"), "error found in agent output!"
