We’ve been getting the metadata overhead warning:
Metadata overhead warning. Over 50% of RAM allocated to bucket "bucket-name" on node "node-name" is taken up by keys and metadata.
This means we have a lot of small documents, yes. But we wanted to silence this alarm for now, since when we monitor our alerts using a call to: /pools/nodes We keep getting these alerts on 2 of our five nodes in the cluster. We’ve tried silencing it with:
curl -sS --data “ns_config:set(alert_limits, [{max_overhead_perc, 75}]).” -u USER:PASSWORD http://localhost:8091/diag/eval
checking it with a call to the api shows the overhead percentage was in fact increased to 75%:
curl -sS --data “{value, Config} = ns_config:search(alert_limits),proplists:get_value(max_overhead_perc, Config).” -u USER:PASSWORD http://localhost:8091/diag/eval
but this alarm still exists. Making a call to alertsSilenceURL also didn’t help.
Is there any way of silencing these alarms until we can fix the underlying issue?