Currently running a 22 node cluster and we are monitoring Bucket stats with the REST API. The vb_active_num value for <30 seconds is lower than the normal 1024 value (returning 967). Any ideas what could be causing this? No nodes are returning a status other than healthy during this short time period.
At a guess perhaps the cluster manager is having difficulties gathering stats from one (or more) nodes? Note that the value for vb_active_num is an aggregation from all the nodes in the cluster.