An availability group keeps your servers from sharing a single point of failure. Put several VPS in a group and CubePath places each one on a different physical host, so if one machine goes down, the rest keep running.
A host here is one of CubePath's hypervisors — a physical server in the datacenter running many virtual machines at once. A group of three web servers lands on three separate hypervisors: if one fails or reboots for maintenance, you lose one server while the other two keep serving.
It's the standard setup for anything that has to stay up: load-balanced web tiers, database replicas, and any service where two servers on the same host would defeat the point of having two.
Create a group
- 1Open the create flowOn the Availability Groups page in my.cubepath.com, click Create Group.
- 2Name it and pick a locationEnter a name (and an optional description), then choose the location.
- 3CreateConfirm. The group is ready to deploy machines into.
Add machines
Pick the availability group when you deploy a VPS, so it's placed correctly from the start. You can also add an existing machine to a group later.
The machine must be in the same project and location as the group, and in a state where it can be modified (not mid-deploy or mid-deletion). Removing a machine from a group leaves it running; it just stops being part of the spread.
ImportantAdding an existing, running VPS to a group doesn't move it. The spread is applied when machines are placed, so a server already running on a shared host stays there. To actually get the separation, deploy the machine into the group from the start.
See the group together
A group view lists its members and shows combined metrics for the whole tier in one place. You can also point a Cloud Alert at an availability group rather than at a single machine, which is the right granularity for "is this tier healthy?".
Limits
| Thing | Limit |
|---|---|
| VPS per group | 50 |
| Scope | One project and one location per group |
What it doesn't do
- It's placement, not load balancing. It only decides which hypervisor each VPS runs on. To share requests between the members, put a Load Balancer in front.
- It's single-location. A group spreads across hosts within one location. For resilience across regions, run a group in each location and route between them with GeoDNS.
TipAn availability group and a load balancer are the two halves of the same idea — the group makes sure your servers can't fail together, the load balancer makes sure traffic stops going to whichever one does.