Closed dry4ng closed 8 months ago
Can you post a log of what you're seeing?
Here is a typical result from the exporter. As you can see most host/task/queue combination generate 770 metrics. While celery_task_runtime_bucket
returns 11205 metrics. And the total number of rows is 25956. In my podmonitoring config (servicemonitor) I drop most metrics except few. I would like to have the dropped at the exporter level.
Why don't u adjust the buckets? Less buckets less metrics. Obv a drawback shortterm by adjusting bucket sizes.
export CE_BUCKETS=1,10,60,600,1800
U can also set CE_GENERIC_HOSTNAME_TASK_SENT_METRIC=true
to set a generic hostname for the celery_task_sent_total
metric. Otherwise it's the host that sent the task, e.g a random django pod which generates label cardinality.
Thanks, that worked.
Good guy Adin
I'm able to limit what is scraped via servicemonitor, but is there a way to limit what is exposed at the /metrics endpoint? Our current setup returns 26k metrics.