Windows Server TechCenter >
Windows Server Forums
>
Windows Server 2008 R2 Hyper-V
>
What about VM's memory when active host down unplannedly
What about VM's memory when active host down unplannedly
- Hi,
I'm trying to undestand live migration process when active node of cluster has a hardware problem and down unplannedly. I read that, there is no VM downtime when a node fails. I think a case like that, there are some VMs on active node and they are running and they are also using memory of host machine. If active node has a powersupply problem and become down, pasive node will start to run using VMs's disks on the shared storage, right? I'm ok for this. But I didn't understand that, what about VMs's memory? All memory data on active node deleted when it's shotdown unplannedly. How this VMs continiues to work without old memory data without downtime?
Thank you
Answers
- High Availability and Live Migration are two distinct features - with different purposes.
The one thing they have in common is that they are managed by Clustering.
High Availability is the act of keeping a machine or application running. If a host goes down (crash) unexpectedly, it is the job of Failover Clustering to recognize that and power up the VM on a different host.
hard crash = potential data loss (memory that is not written out, there is no opportunity to save the memory)
(this is what Clustering has been doing for its entire existence (many years now) - manage services and keep them running, on one of the hosts in the cluster)
Live Migration is the act of moving the running workload to a different host. In this case, there is no hard crash. Thus the running menory state is transferred and nothing is lost.
no crash = no data loss.
This is the same no matter who's hypervisor you are using.
the only way to prevent data loss is to have constant writes to disk, this can also end in small data loss in the event of a crash - no different than if the application was running on hardware and you trip over the plug.
All the same rules apply.
Brian Ehlert (hopefully you have found this useful)- Marked As Answer byVincent HuMSFT, ModeratorMonday, September 21, 2009 6:23 AM
- Proposed As Answer byVincent HuMSFT, ModeratorThursday, September 17, 2009 9:22 AM
All Replies
- High Availability and Live Migration are two distinct features - with different purposes.
The one thing they have in common is that they are managed by Clustering.
High Availability is the act of keeping a machine or application running. If a host goes down (crash) unexpectedly, it is the job of Failover Clustering to recognize that and power up the VM on a different host.
hard crash = potential data loss (memory that is not written out, there is no opportunity to save the memory)
(this is what Clustering has been doing for its entire existence (many years now) - manage services and keep them running, on one of the hosts in the cluster)
Live Migration is the act of moving the running workload to a different host. In this case, there is no hard crash. Thus the running menory state is transferred and nothing is lost.
no crash = no data loss.
This is the same no matter who's hypervisor you are using.
the only way to prevent data loss is to have constant writes to disk, this can also end in small data loss in the event of a crash - no different than if the application was running on hardware and you trip over the plug.
All the same rules apply.
Brian Ehlert (hopefully you have found this useful)- Marked As Answer byVincent HuMSFT, ModeratorMonday, September 21, 2009 6:23 AM
- Proposed As Answer byVincent HuMSFT, ModeratorThursday, September 17, 2009 9:22 AM
