I had some node restarts and so we can see cluster size was changing
However if I zoom out to about last 24 hours it looks like Cluster did NOT
ave any times it operated in reduced size
The problem is we're not using min_over_time to match the selected
$interval as such if the time when node left cluster was short it could not
be picked up with gauge "sampling" which happens.
Peter Zaitsev, CEO, Percona
Percona Live 2018 Open Source Database Conference
<https://www.percona.com/live/18/> April 23 - 25, 2018 Santa Clara,CA Get
Your Ticket Now!, Inquire about Sponsorship!