This whitepaper is for historical reference only. Some content might be outdated and some links might not be available.
Summary
The following table summarizes the pros and cons of the evacuation patterns described.
Table 5: Evacuation pattern pros and cons
| Approach | Pros | Cons |
|---|---|---|
| Data plane-controlled evacuation |
Relies only on data plane actions Quickly prevents work from being done in the impacted Availability Zone Flexible approach to a centralized view of Availability Zone health |
Does not prevent capacity from being deployed in an impacted Availability Zone Not all workload types can use this approach easily |
| Control plane-controlled evacuation |
Prevents new capacity from being deployed in the impacted Availability Zone Removes existing capacity from the impacted Availability Zone |
Relies on each service’s control plane Requires code to be written for each service Has to be completed service by service Needs to be careful not to overwhelm capacity during the update |
You will likely use both approaches together as part of an Availability Zone evacuation plan. Start with the data plane-controlled evacuation actions that are more likely to succeed to quickly stop processing work in the impacted Availability Zone. Then, once the initial impact is mitigated, follow-up with the control plane-controlled evacuation actions, if you deem it necessary.
