VMware Cloud Community
bardanes
Contributor
Contributor

High Availability/Failover with Free Hypervisor

Hi,

Our organization is new to virtualization and will be deploying the following configuration:

Two servers: primary VMWare Server (to be purchased, running 10 VMs (mostly RHEL))  and backup server (already purchased), which will come online only if failure/problems with primary.

iSCSI SAN for shared storage. Backups performed to Buffalo NAS appliance.

We have been using the free vSphere Hypervisor in testing with the VMs on the soon-to-be backup server.

We are well-aware of the advantages of the licensed version and vCenter Server.

However, we would like to continue to use the free Hypervisor and are trying to develop a high availablity/failover strategy with the backup server. I state this relatively, because this is not a time-critical environment that requires immediate availability and zero downtime.

Our current exploratory strategy:

1. Primary server goes down.

2. Start secondary server.

3. Access resource manager and login.

4. Search for and remove any locks on file system datastores (iSCSI reservations?).

5. open VMWare Client.

6. Search for and instantiate VMs that were previously managed on the primary server from shared storage.

We have not tested this and are making a number of assumptions about this process.

Does anyone have experience with this with the free Hypervisor?

Thanks,

8 Replies
wealvescabral
Enthusiast
Enthusiast

Hi,

You will need a vCenter Server installed, create a new cluster, enable HA and put the hosts on it.

You can test vCenter Server for 60 trial license.

Explanation about HA on vmware:

http://www.vmware.com/files/pdf/VMware-High-Availability-DS-EN.pdf

Also there are another tools that make VM replication like Veem Solution.

Wellington Cabral | Mark it as helpful or correct if my suggestion is useful.
0 Kudos
bardanes
Contributor
Contributor

Thanks for your reply - we are trying to do this without vCenter Server (and paid licenses) if possible, just the free Hypervisor.

0 Kudos
tomlee
Contributor
Contributor

Hi,

i think alternative is using NFS, you could share path/datastore to both servers, backup server for some/security.. reason can has read only access.

if primary server goes down you can simply change permissions to RW and power on vm's

of course i dont know yours Buffalo storage, i did not read anything about it.

it is just my concept, i enjoy when it helps you a little bit

Thanks & kind regards,

tomlee

0 Kudos
bardanes
Contributor
Contributor

Thank you for your reply. This was originally along the lines of our thinking. However, the backup server will only come up in a failover circumstance - we cannot run the same virtual machines on two free hosts simultaneously - there is only low-level management of each host/shared storage that has been provisioned and also no clustering - each host does not know about the other. The VMs on the shared storage should be locked by the primary server.

If and when the backup server is brought up, the datastore will be scanned by VMWare Client and the VMs will be hopefully be inventoried and powered on for the first time. In the best of all worlds, there is no data loss, locks are relinquished, and the scanning process won't take more than a few hours. Then, the virtual machines can inventoried and powered on and we are back to where we were.

We just don't know about the state of the VMs on the shared storage in a failover scenario - the vCenter tools mask the filesystem and perform the recovery. The backup host would then be part of a cluster and with replication, will start up immediately based upon the rules of the vCenter Server before it dies.

The question again (somewhat rephrased) is - in a simple configuration without time-critical needs for failover and high availability and automated backup through ghettoVCB scripts, is it necessary to have the additional tools that clustering, replication, and backup agents provide?

I am just trying to determine if anyone has attempted what we have described earlier and what they've found.

0 Kudos
tomlee
Contributor
Contributor

In my opinion you don't need any extra tools. If you will do it in way like you mentioned, it will be work.

But you have to do all things manually. I think you will spend a lot of time when something goes wrong.

Less work is better, be smart lazy :-), that I suggest use both servers (each with his own datastore) for hosting vm's. If there will occur any problem with one server, second will still working. Only half number of vm's will be inaccessible.

I prefer use NFS, from my expierence I had much less troubles with that than with iscsi, but maybe I know it not enough.

tomlee

0 Kudos
bardanes
Contributor
Contributor

Thanks again for your reply.

How are you using NFS for your shared storage? Are you running a virtualized NFS with something like FreeNAS or an appliance? Where is your shared storage, on DAS drives or on the network?

The issue for us with the shared storage solution also is redundancy and/or failover if it goes down. I will probably post this as a separate question.

0 Kudos
bccahill
Contributor
Contributor

If you can use another product beside ESXi, you could try XenServer.  It includes XenMotion with the free edition.  I've used it in smaller environments like yours and its worked well.

http://www.citrix.com/products/xenserver/features/editions.html

0 Kudos
cyberpaul
Enthusiast
Enthusiast

Sorry to resurrect an old thread, but I just came across this while looking for something similar.

This scenario - using vSphere cluster for primary site and leveraging ESXi Free for disaster recovery - makes a perfect sense to me.

I use VMcom backup for this (https://vmcom.com). You set up a primary site as you would normally do. Set up VMcom for backing up the primary site as you would normally do. Then you utilize the "Backup Copies" feature to periodically copy the latest restore points to an ESXi Free host using SCP. It even allows you to store multiple restore points so you can immediately start an older backup if you wish.

Ideally, your ESXi Free host would have enough disk space to serve as a secondary backup storage under normal conditions. In case your primary cluster goes down bad, you can immediately start your VMs on the backup host. Just browse the datastore for the restore point you wish and power-on selected VMs.

Personally, I even use VMcom's Backup Copies feature to upload some VMs to my office Mac with VMware Fusion. Before I work on production VMs, I can safely test everything on a recent dataset.

Merry Christmas!