Mozilla Telemetry Histogram Analysis Using Pearson's Correlation Coefficient
Analyzes each histogram for any change in behavior.
1. Sample Configuration
filename = "moz_telemetry_histogram_monitor.lua"
ticker_interval = 600
preserve_data = true
message_matcher = "Uuid < '\003' && Fields[normalizedChannel] == 'release' && Fields[docType] == 'main' && Fields[appVendor] == 'Mozilla' && Fields[payload.histograms] != NIL" -- slightly greater than a 1% sample
histogram_samples = 25 -- number of histograms to compare 24 historical + current interval
histogram_interval = 3600 -- collection period for each histogram
alert = {
disabled = false,
prefix = true,
throttle = histogram_interval,
modules = {
email = {recipients = {"trink@mozilla.com"}},
},
thresholds = {
-- pcc = 0.3, -- default minimum correlation coefficient (less than or equal alerts)
-- submissions = 1000, -- default minimum number of submissions before alerting in at least the current and one previous interval
-- ignore = {} -- hash of histograms to ignore e.g. 'histogram_name = true'
active = histogram_interval * 5, -- number of seconds after histogram creation before alerting
}
}
source code: moz_telemetry_histogram_monitor.lua