Todays learning: VMware HA - need to buy more ram

A

AMD_Gamer

Fully [H]
Joined
Jan 20, 2002
Messages
18,287
After getting my napp-it ZFS all-in-one up and running for a sweet shared datastore i spent tonight learning HA. Everything seems to be working great but 8GB of ram per host is not going to cut it. I had all kinds of trouble getting HA to work and even start up VM's in the cluster until i selected the Admission Control and let the VM's violate the resource constraints.

I tested it it by disconnecting the network cable, and doing a hard power down of the host via the PSU switch.

Does anyone know a good way to shutdown an individual VM? like make it BSOD or something? i tried putting them on a seperate vSwitch with its own NIC and the heart beat still comes through the other vSwitch. Is HA commonly used with individual VM's or is it more for the host level?

2abnf.jpg


Uk324.jpg
 
8gb is still alot of ram. how much did u set a side for your zfs storage?

are you building a dedicated storage box anytime s00n? doo eeettt :D
 
Yeah..what is your RAM allocation on those VMs? Go research vSphere HA "slot sizes". :)
 
The ZFS VM is 4GB and I bumped the Server 2008 VM's down to 2GB.

Dedicated NAS is next on the agenda once I get the funds. I don't really like the all in one solution as it is not something you would do in production. I mainly did it as a learning experience for pass-through and i was eager to learn about ZFS and play around with napp-it, FreeNAS 8 and ZFSGURU.
 
If you have multiple NICs and different VM's attached to a certain NIC, how do you get HA to work when any NIC gets unattached? I setup a new vSwitch with a dedicated NIC and set a VM to use that. I then disconnected the network cable but HA did not do anything because the heartbeat is going out from the other vSwitch?:confused:
 
AMD_Gamer said:
If you have multiple NICs and different VM's attached to a certain NIC, how do you get HA to work when any NIC gets unattached? I setup a new vSwitch with a dedicated NIC and set a VM to use that. I then disconnected the network cable but HA did not do anything because the heartbeat is going out from the other vSwitch?:confused:
Click to expand...

HA is only designed to respond when a host becomes isolated, or becomes entirely unavailable (ie. fails). Removing a cable from a single NIC, unless as you say, it is the last NIC in the team that heartbeat traffic is traversing, will not do anything.

Isolation being when a host cannot communicate with any other HA primaries (via VMK ports on ESXi or the SC on ESX), and cannot reach the isolation address, however is able to communicate with storage. This is where your isolation response setting comes in to either safely clear the file locks on storage (via powering off the VM or shutting it down) or maintaining status quo ("do nothing").

Host failure looks the same to the "observing" HA primaries, however the storage locks should clear quickly and will enable the guests to be powered up quickly by HA.
 
Yep. HA is there to protect against a host failure or when a VM stops sending heart beats via tools...and eventually it'll also monitor application heart beats. You really never have a single NIC in a port-group..you always team them to provide redundancy, preferably to a second switch so there is no SPoF.
 
Yep. HA is there to protect against a host failure or when a VM stops sending heart beats via tools...and eventually it'll also monitor application heart beats. You really never have a single NIC in a port-group..you always team them to provide redundancy, preferably to a second switch so there is no SPoF.
 
Then what about the VM Monitoring tab? They have a whole tab for individual VM's.
 
You must log in or register to reply here.
Back
Top