opmon/telemetry-alerts-prototype.toml

[project]

name = "Telemetry Alerts Prototype"
platform = "firefox_desktop"
xaxis = "build_id"
compact_visualization = true
skip_default_metrics = true
start_date = "2022-06-01"
alerts = [
    "hist_diff",
    "crash_thresholds"
]
metrics = [
    "checkerboard_severity",
    "gc_ms",
    "main_crashes",
    "oom_crashes",
    "js_pageload_execution_ms",
    "memory_total",
    "gc_ms_content",
    "gc_budget_overrun",
    "perf_first_contentful_paint_ms",
    "active_ticks",
    "subsession_length"
]

[project.population]

data_source = "main"
monitor_entire_population = true
dimensions = ["os"]
channel = "nightly"


[metrics.main_crashes]
data_source = "crash"
select_expression = "SUM(IF(payload.process_type = 'main' OR payload.process_type IS NULL, 1, 0))"

[metrics.main_crashes.statistics]
sum = {}

[metrics.oom_crashes]
data_source = "crash"
select_expression = "SUM(IF(payload.metadata.oom_allocation_size IS NOT NULL, 1, 0))"

[metrics.oom_crashes.statistics]
sum = {}

[data_sources]

[data_sources.main]
from_expression = "mozdata.telemetry.main_nightly"
submission_date_column = "DATE(submission_timestamp)"

[alerts]

[alerts.crash_thresholds]
type = "threshold"
metrics = [
    "main_crashes",
    "oom_crashes"
]
min = [0]
max = [500]

[alerts.hist_diff]
type = "avg_diff"
metrics = [
    "checkerboard_severity",
    "gc_ms",
    "js_pageload_execution_ms",
    "memory_total",
    "gc_ms_content",
    "gc_budget_overrun",
    "perf_first_contentful_paint_ms",
    "active_ticks",
    "subsession_length"
]
window_size = 7
max_relative_change = 0.25
percentiles = [25, 50, 75, 95]