ESXI failover of VMs
- Thread starter GodAtum
- Start date
Associate
- Joined
- 3 Oct 2007
- Posts
- 795
VSphere Fault Tolerance will do what you describe, it creates an exact in step mirror that takes over if the 'live' ESXi host fails.
Anything that happens on the Live box happens on the mirror, so this is no protection against a crash, reboot, etc - It's only protection against the underlying ESXI box failing.
You need to have the appropriate licensing: Standard or Enterprise for up to 2 CPU's, Enterprise Plus for up to 4 CPU's. A 10Gb network between the two hosts is highly recommended, as is shared storage though if you're using VSAN this will work too.
Needless to say it's quite a list of requirements and is massive overkill in nearly all circumstances and HA is generally good enough.
Anything that happens on the Live box happens on the mirror, so this is no protection against a crash, reboot, etc - It's only protection against the underlying ESXI box failing.
You need to have the appropriate licensing: Standard or Enterprise for up to 2 CPU's, Enterprise Plus for up to 4 CPU's. A 10Gb network between the two hosts is highly recommended, as is shared storage though if you're using VSAN this will work too.
Needless to say it's quite a list of requirements and is massive overkill in nearly all circumstances and HA is generally good enough.
Associate
A second option still has the storage shared between ESX hosts, but if one host dies then the second ESX host boots the VM. So the time it takes to failover is the time the VM takes to boot on the second ESX host. Your application hosted on your VM will have to auto start each time the VM boots for it to be relatively seamless.
The important feature is that the ESX DataStores (storage) are centralised on a SAN or NAS device. You need to build in redundancy into the SAN or NAS as well! The application requirements will determine IOPs the SAN/NAS needs to sustain, it could be low and a 1Gbit link to the storage may be adequate, especially if you mitigate memory paging to the SAN. By having a local SSD drive for ESX page memory or reserving the VM guest memory to eliminate memory paging.
As said, you will have to licence ESX for DRS Failover. Or you could use a Linux VM to monitor and execute scripts remotely (SSH) to bring the VM on the second host online, saving £1000s in licensing costs. Just search www and you will find examples on how to do this!
The important feature is that the ESX DataStores (storage) are centralised on a SAN or NAS device. You need to build in redundancy into the SAN or NAS as well! The application requirements will determine IOPs the SAN/NAS needs to sustain, it could be low and a 1Gbit link to the storage may be adequate, especially if you mitigate memory paging to the SAN. By having a local SSD drive for ESX page memory or reserving the VM guest memory to eliminate memory paging.
As said, you will have to licence ESX for DRS Failover. Or you could use a Linux VM to monitor and execute scripts remotely (SSH) to bring the VM on the second host online, saving £1000s in licensing costs. Just search www and you will find examples on how to do this!
Fault Tolerance is a pretty horrible caveat-filled workaround for poorly written/legacy services that don't handle failover or clustering at the application level.
HA will boot VMs on a different host in the event of a host failure, providing the datastores are accessible to the new host.
I think you need to take a few steps back - what problem are you trying to solve?
HA will boot VMs on a different host in the event of a host failure, providing the datastores are accessible to the new host.
I think you need to take a few steps back - what problem are you trying to solve?