Vsphere High Availability 2 hosts

[Avdaga]

Dear all,

I am trying to achieve the following using the essentials plus package. I have got 2 identical machines each with 2 tb storage. Both are running ESXI and I want to setup High availability and load balencing. From what I read on the fora's is that vmware requires shared storage so some sort of san if I want to use High availability. However I do not have this machine, what I want to achieve is the following: Vmware mirrors the disks of both machines (network raid 1) and if a hardware failure occurs on one of them the other machine will boot all of the remaining vm's. In the same way vmware should do load balancing. From what I read on the internet vmware had VSA however this was discontinued. VSAN is not an option since I do not have 3 machines. A possible option would for example be starwind however I would prefer an option from vmware itself.

So in short is there a way to setup High availability and Load balancing with 2 hosts and no shared storage? Preferably without third party software.

Kind Regards,

Sebastian Wehkamp

[Wscholz]

Hmm,

i see a lot of problems in your setup. If you loose a host in your scenario you have two different desasters, one is a host-  and the other one is a storage malfunction. So the question is, how do you get your vm´s replicated (including vcenter i guess) and the other one how do you somehow emulate the intelligence to check whether a host is really broken. Have you looked at the VSphere Replication appliance ? Maybe you can play around a litte bit with that. There is also a nice article here

VMware Front Experience: Automating VM recovery with stand-alone vSphere Replication

about automating recovery. Regarding load balancing, i don´t think that you can automate something like drs does, but you could always decide manually what vm you deploy where and only during a host failure the remaining host has to serve all the vm´s. Overall i would be really careful in that scenario not to do things that are unsupported. The starwind solution you mentioned looks nice and to be honest it seems to me worth to try, might be less headache than the replication appliance.

[Avdaga]

Thank you for your reply. I figured that since both of the storages are exactly the same the secondary machine would be able to easily boot the lost vm's. For the syncing I thought there would be something like drbd, when the hosts cannot connect with each other secondary will become primary. This will work in the case of a hardware failure and in case of a network failure a split brain will occur which must be resolved manually. I will play around with the replicaton appliance and reconsider the starwind software.

[Anton_Kolomyeyt]

Technically you can use Linux and DRBD (or FreeBSD and HAST if you're on the dark side of the Moon) to create a fault-tolerant block device cross-replicated between a pair of VMs running on a different hypervisor nodes. Throw in

failover NFSv4 mount point on top of it and you have a nice VMware VM datastore. Tons of a Software Defined Storage providers do have exactly this ideology inside their Virtual Storage Appliances so you're not going to be different :smileyhappy:

However as StarWind Virtual SAN is FREE for 2-node setup within VMware scenario (Hyper-V licensing is different if you care) and can run on top of a free Hyper-V Server (no need to pay for Windows licenses) you MAY end with a faster

turn-around going StarWind route. Depends on what you want from a features set (iSCSI? NFS? SMB3? In-line dedupe? RAM cache?) and what forum you prefer for a public support to ask questions :smileyhappy:

--

Thank you for your reply. I figured that since both of the storages are exactly the same the secondary machine would be able to easily boot the lost vm's. For the syncing I thought there would be something like drbd, when the hosts cannot connect with each other secondary will become primary. This will work in the case of a hardware failure and in case of a network failure a split brain will occur which must be resolved manually. I will play around with the replicaton appliance and reconsider the starwind software.

[Avdaga]

Thank you for your reply, this answers my questions :smileyhappy:

[AnatolyVilchins]

Hi! Just want to share some key benefits of the solution here (jfyiyou know):

1) Minimalistic hardware footprint

a. Start as small as only two nodes (“TWO” means “TWO” and not “TWO and some witness node or voting service in cloudor somewhere or whatever” like other VSA guys do) and Scale-Out to infinity J

b. Gigabit Ethernet (we’ll take care of 10 GbE or 40 GbE of course, but lack of them is not a show stopper)

c. Switch-less point-to-point configuration (fully supported and approved for production)

d. Flash is optional (would be used as a write-back cache or as a tier in an all-flash configs if installed, but no flash is no big deal)

2) Flexed out “VMware vSphere friendly” licensing

Customer pays for actual consumed storage (by number of nodes, orby used terabytes – he decides on his own) independently fromamount of hypervisor hosts in a cluster. It’s perfectly OK to have StarWind Virtual SAN installed on just a few nodes in a much bigger cluster – there’s absolutely no need to enable it and pay license fees for all of the nodes.

(For example: 5 node vSphere cluster can have VM storage on just two nodes, 8 hypervisor nodes can run VMs from three node storage cluster and so on)

Two-node VMware vSphere setup is FREE of charge. Capacity-unlimited, perpetual licenses, production use allowed. Technically you can run your whole datacenter from just a pair of StarWind Virtual SAN nodes for free. As long as you don’t run out of capacity and IOPS, and network bandwidth of course :smileyhappy:

3) All-in-Wonder” ©... virtual storage solution :smileyhappy:

Industry-standard storage protocols exposed (iSCSI, SMB3 and NFSv4), in-line 4KB block deduplication, asynchronous replication, RAM and flash write-back caches, log-structuring etc.

There’s no point to spawn a set of a failover NFS server VMs on top of a StarWind Virtual SAN just because customer needs a place to store VDI profiles or he needs to have a company SMB3 file server – we do this out of the box. Obviously other storage consumers outside the cluster are also supported because of that including “alien” hypervisors like Hyper-V, Xen and KVM.

There’s no need to deploy any flash and RAM caching software as StarWind Virtual SAN comes with own caching engine. In-line deduplication combined with a log-structured file system takes good care of a performance targeted all-flash and capacity-oriented parity RAID configurations. Roll-your-own-NetApp

[DinkyChikka]

VM Replication could be used.

Sent from my iPad