Why won't my virtual machine restart after the host fails?
Modern hypervisors like VMware ESXi offer powerful and reliable platforms for virtualization. Even though the hypervisor’s stability and performance are constantly improving, there are many situations when virtualization features fail or behave unexpectedly. When trouble strikes, IT administrators need to understand the problem’s root cause and take decisive action to correct it and minimize any adverse impact on the business.
Tools like vSphere High Availability (HA) are intended to restart failed virtual machines (VMs) when a failure occurs on a host system or within a VM itself. If the VM fails to restart in a timely manner, the problem can usually be tied to a problem with HA and its configuration. However, always begin a troubleshooting investigation by verifying that the HA tool is actually protecting the VM -- if not, the VM will not restart and you’ll need to try to restart the VM manually. Make sure that tools like HA are working as soon as the VM is running.
Next, examine the available computing capacity on each potential host server. Verify that any potential destination server runs a compatible hypervisor -- vSphere VMs may not run on Hyper-V systems. Also make sure that adequate computing capacity is available to actually restart the afflicted VM. Be aware that many VMs will require slightly more resources to startup. This means servers with lean resources may experience startup glitches without a bit of additional workload balancing or resource adjustment.
When problems persist, consider the possibility of VM file damage or corruption. It may be necessary to restore a troubled VM from a backup and attempt recovery from snapshots or other data protection methodologies.
Prevent VMware High Availability problems before they start
VMware High Availability updates in vSphere 5
Understanding the difference between VMware High Availability and Fault Tolerance
How does VMware Fault Tolerance work?
Dig Deeper on Virtual machine monitoring, troubleshooting and alerting
Related Q&A from Stephen J. Bigelow
Just because software passes functional tests doesn't mean it works. Dig into stress, load, endurance and other performance tests, and their ... Continue Reading
Don't neglect form factor as part of your data center server selection. Instead, figure out what type of environment you need and learn which server ... Continue Reading
Learn how load balancing in the cloud differs from a traditional network traffic distribution, and explore the different services available from AWS,... Continue Reading