Virtual machine HA:
The virtualized environment uses shared storage. When the cloud platform detects a computing node failure or a virtual machine failure, the system automatically starts the failed virtual machine on the normal computing node to make it recover quickly. The key to HA is the disk lock technology to avoid brain splitting.
HA trigger condition:
When the server is faulty, HA is performed.
When the virtual machine fails, HA is performed.
The host recovery time is 80-205 seconds. The subdivision steps are as follows:
1. CPS detection failure time: 30s
2. Ceilometer detects a fault: 0-60s
3. Ceilometer issues an alarm: 0-60s
4. HA waiting time: 5s
5. Nova determines brain split: 5s
6. VM reconstruction: 15s
7. OS loading: 30s
