Interesting feature request? :)

[pkelly_sts]

I've just completed a migration to new hosts/storage and tidying things up whilst also monitoring performance of the new storage.

Our proxy VMs are largely inactive during the day whereas I noticed them touching the storage occasionally (whilst monitoring latency) and these are used alongside a direct-attached repo/proxy physical server to improve backup bandwidth during jobs.

It occurred to me that these are reasonably highly spec VMs (cores/memory) on the basis that in our environment at least most production VMs aren't doing much at night so these VMs have a completely opposite usage pattern to the production VMs.

I then thought how it could be useful to have the proxy VMs powered off by default (except perhaps keeping one powered on for minimal daytime tasks) but realised that it could be useful to perhaps spin them up on demand, thus only consuming resources as & when actually required.

Petty/tight-fisted or useful idea?

[foggy]

Hi Paul, thanks for sharing - I recall this being mentioned once or maybe twice before and someone tried to implement this via scripted shutdowns/bootups. Sounds nice but demand is considerably low.

[Andreas Neufert]

This approach for all kind of components including Repositories make perfect sense when they run in the cloud like AWS EC2. Potentially you could save half the money there.
The more we will go into the cloud the more likely is maybe such a usecase. Let´s see.

[Andreas Neufert]

For your workload. Yes the VMs have a lot of allocated ressources but by default the VM do not use it and can be consumed by your other VMs.

[pkelly_sts]

Yeah I can understand demand for internal workloads being low, but can definitely see it for cloud-based. Thanks for the comments