I wonder if anybody else bumped into this issue?
- You get a couple of beefy 12-core VPS instances in some cloud like DigitalOcean and set up a database on them in "active-passive" cluster configuration.
- Everything works great initially, you are saving a lot of money compared to dedicated servers, and high availability is there.
- If you don't failover this cluster for a long time, your cloud provider will notice that your Standby server never uses the 12 CPUs. And they will add a bunch of other VPS instances to your host machine to use up all underutilized capacity.
- Then your next failover causes a major disaster because your host server gets overwhelmed, and your database is unable to handle the normal load.
Do you know of any workarounds for this issue that don't cost a fortune? I am thinking about scheduling some script that would check the recent system load and create some artificial workload just to make sure that the hypervisor does not underestimate the resource requirements of Standby servers... Do you have a more elegant option?
