We use application-only monitoring on a managed OpenShift PaaS and recently I noticed that each of our dead container instances (showing up as "hosts" in unmonitored state) still happily consume host units even after they've been long restarted/terminated (it seems up to 72 hours). In our case this can be verified easily in the "Deployment state" view:
Each of the above "hosts" belongs to a terminated container. We have LOTS more of these and the total HU consumption is hugely significant (in our case it is almost 25% of our total HU volume).
If I manually hit "Disable monitoring" for these then they will no longer contribute to the HU consumption number.
Questions that come to mind:
We are currently on Managed 1.174.
Any feedback is welcome.
It's difficult to compare the HU consumption number in the deployemnt status with the one in the CMC because we have more than 400 hosts in our environment and simple filtering with Status=Unmonitored over all hosts is not possible (it's always limited to 400 hosts). Plus, there is a lot of fluctuation throughout the day and - if I'm not mistaken - the figure in the CMC is calculated once every 24h.
But using different filters in the deplyoment status view and adding the HU's manually I'm getting pretty close to the number in the CMC... so yes, I would say I'm pretty sure they are counted.