In order to keep the size and cost of the proxied_request_metrics table from growing infinitely, and to provide useful aggregated metrics for endpoint operators to analyze traffic patterns, the proxy service runs several processes to routinely compact that table
The Metric Pruning routine is responsible for deleting any metrics that are older than N specified days from the proxied_request_metrics table. It is controlled via the following environment variables
# Whether metric pruning routines should run on the configured interval, defaults to true
METRIC_PRUNING_ENABLED=true
# How frequently metric pruning routines should run
# defaults to 1 day
METRIC_PRUNING_ROUTINE_INTERVAL_SECONDS=10
# how long (after the proxy service starts) it will wait
# to run the first iteration of the metric pruning routine
METRIC_PRUNING_ROUTINE_DELAY_FIRST_RUN_SECONDS=5
# The maximum number of days of request metrics
# that should be preserved measured
# from the time the pruning routine runs
# all metrics for earlier days will be truncated
METRIC_PRUNING_MAX_REQUEST_METRICS_HISTORY_DAYS=45#TODO