Vmware Vsphere HA ( Shared Storage Problem ) and Network Bandwith Problem

[coolirc]

Hello All

i'm setting up VMware VSphere HA on ESXI 6.0

i have 5 hosts currently running, each host have 2 management networks , 1 vMotion network , and 1 nic load balanced with the main management network . ( each nic 1 Gbit uplink )

i also attached to each host an FC card ( Fiber Channel ) connected to HP 3PAR Storage .

each host have 1 TB or 2 TB Fiber Channel attached storage .

when i was configuring the HA it stated that no shared storage was available how can i allow other hosts to see each datastore as a shared storage . NB : when i go vsphere web client i can see all the storage of each host .

also i want to enable FT on a vm i have also a problem with insuffisent bandwith or something . .

also i've enabled fault tolerance on each nic attached to the hosts .

also one more question : should i export the volumes ( luns ) to all the hosts in the HA ? ( actually  i exported each volume to each host separately ) for example :

Volume 1 : HP3PAR_ESX1 exported to HOST ESX1

Volume 2 : HP3PAR_ESX2 exported to HOST ESX2

Volume 3 : HP3PAR_ESX3 exported to HOST ESX3

...etc

can someone clear me about the issue or guide me to where i can solve those problems thanks .

when i try to activate the FT on the VM i have the following warning attached

Ce message a été modifié par : coolirc attached FT image problem

[npadmani]

also one more question : should i export the volumes ( luns ) to all the hosts in the HA ? ( actually  i exported each volume to each host separately ) for example :

Volume 1 : HP3PAR_ESX1 exported to HOST ESX1

Volume 2 : HP3PAR_ESX2 exported to HOST ESX2

Volume 3 : HP3PAR_ESX3 exported to HOST ESX3

...etc

Shared Datastore is mandatory requirement for HA.

Yes, you got to export those all LUNs to all the hosts so they become shared LUNs, that way they become shared VMFS datastores.

in order to enable FT for your VM(s), you must have vSphere HA cluster up and running.

Now warning that you are facing, I wasn't able to read it, but if you have 1 Gbps network link for FT, it will give you warning that have at least 10 Gbps network bandwidth.

[coolirc]

thanks for your help still the network bandwith problem i don't understand since i have a 1 gbit nic with fault tolerance activated on it .

[npadmani]

for vSphere FT, 10 Gbps is recommended network bandwidth that we need.

it will work on 1 Gbps too but will be extremely slow and as far as I am aware, in production it's not supported.

[coolirc]

thanks for your reply

is it possible to use the FC Port used for storage for the FT ? i think it will be more handy, i tried to activate the fault tolerance on vm with 10 GB of ram et 300 GB of disk but it says insufficient resources to satisfy the failover level ( translated from french )  ( Ressources insuffisantes pour satisfaire le niveau de basculement configuré de vSphere HA.) even if the remote host has no vm on it and is 12 gb of ram and another one is 15 gb of ram . but still it doesn't allow me to activate the FT .

NB : on the network tab i can only see vmnics on 1 Gb attached to the server

[npadmani]

If you have VM with 10 GB RAM allocation, and 300 GB virtual disk (is this thin/thick?) attached to it.

when you turn on FT for VM, system will raise memory reservation for the VM to 10 GB, in short it's going to make your VM's RAM reservation same as allocation automatically.

Mimimum resource requirements to power on a VM = VMKernel Overhead (usually this is not too big, but depends upton size of VM) + reservation

now let's take your ESXi hosts, one of those host has 12 GB RAM and another one has 15 GB ram installed in it.

Let's pickup 12 GB RAM host and power it on, VMKenel will use some resources out of it, it will also keep some resources as VMKernel system reservation so (in my experience) I have seen roughly you will lose about 3 GB or little more. and you will have somewhere around 9 GB or little less to be used.

in this situation even if you have that one VM with 10 GB RAM allocation, it's not going to work because it's not meeting minimum requirement to power on that VM. In your case i believe you tried to enable FT while VM is powered on already, and the failure message comes up.

plus, FT (in vpshere 6) is also going to create your VMDK files replica, so primary and secondary FT machines will have their independent VMDKs, so storage space needs to be also taken into account. In your case your VM has 300 GB vmdk so please have at least that much extra space ready in one of the other shared datastore.  FT in vSphere 6 supports think provisioned disks.

On top of all the above discussion, you got to also consider resources reserved for failover by Admission control policy, even if you disable in your case, your 10 GB RAM FT machine will fail to power on.

Please use following guide

http://pubs.vmware.com/vsphere-60/topic/com.vmware.ICbase/PDF/vsphere-esxi-vcenter-server-60-availab...

Page 45 onwards you have detailed documentation for FT.

=============

Last month I participated in a discussion for FT resource requirements, see following thread where we tried to see how much is the vmkernel overhead requirements for FT apart from VM's own reservation.

FT on vsphere6

=================

now regarding FT Networking, is that FC HBA that you are using to talk to your FC SAN? in that case you cannot use that network for FT. because FT needs ethernet network and your storage network is pure FCP.

Since you said you have all 1 Gbps VMNICs attached in your environment. It will still allow you to turn on FT with a warning for network bandwidth, just that don't use it in production please.

[coolirc]

Hello and Thanks Again

so as i understand after some reading we need at least the double of the ram available for both source and destination host for the needed VM to work with FT + the admission control needs ,

so in my example my VM has a 10 Gb of Ram and the current ESXI Host has 16 GB of ram and 2 VMS of 1 GB ram

and the destination host one of 16 GB and on of 12 GB

for the storage i don't think it will affect because i have plenty of disk space available . the only problem now is the ram . is it possible to use an ssd disk to allocate the missing ram ?

so as i understand will it ask like 20 GB of RAM for the destination host ? is it only the destination host's ram concerned or the source host also is included ?

also for the network we're planning to upgrade our servers wich nic will be good for the esxi hosts for use as FT port group ? ( HP Proliant DL 380 G7 and Gen 8 )

i attached the screens of My Source Host status and destination hosts status . ram etc . also the vm ressources image

[npadmani · Accepted solution]

Memory reservation requirements for FT enabled VM is across both the VMs, primary and secondary. so it will consume same amount of resources from both the ESXi hosts.

VM with 10 GB RAM allocation, turn on FT so VM's reservation is now 10 GB RAM, so 10 GB + VMkernel overhead consumption from host where Primary VM is running and same amount from host running with secondary VM. on top of this admission control to support host failures to tolerate so let's say primary VM host fails, then secondary VM becomes new primary and triggers a new secondary to be created and powered-on in one of the other surviving host in your HA cluster but that will be successful only when you have enough resources available to satisfy reservation.

your source Host has got 16 GB RAM and one of the destination host has got also 16 GB RAM, so if you try to turn on FT after disabling Admission control policy on your cluster, I think it should allow you to enable FT without errors. [if it still fails then if possible, power down all the other VMs to free up resources or try to reduce their reservation manually if they are running with reservation and try again.]

with admission control, resource requirement would be even higher.

screenshot of Cluster summary is reporting that you have 4 hosts in it, and available memory resource about 30 GB out of total 60 GB memory. But this is still not the right figure, if you go to Monitor -> Resource Reservation, that's where you will find actual 'Available Reservation', that's the right amount of memory reservation that you will be able to use across multiple VMs you are running in your cluster.

I have tried 2 vCPU vm FT in vSphere 6, works quite well but never tried it with Hardware version 8, in your case your VM is hardware version 8, although I am not too sure if it's mandatory to have hardware version 11 VM only for FT so ideally it should work, but if possible upgrade vm virtual hardware version to 11.

[coolirc]

Hello Again and thanks for your clarifications

i tested the FT yesterday on my VM by reducing the Ram from 10 Gb to 6 Gb and then activated the FT and it worked , now the original host has 2 vms running with each 1 Gb and the FT vm of 6 Gb and the hosts total use of memory is 10.49 Go .

now let's see the destination FT host wich is a total of 16 Gb of ram , currently using 8.56 Go of total memory , and the vm is using 6144 Mo of ram from the HA cluster Menu when i click on the Monitor Tab Ressource Reservation i see VM(Primary) is using 6144 Mo and also the VM (secondary) .

so conclusion it's using the original VM's memory and the vmkernel's memory ( this with admission control disabled ) .

as for the network i'm currently using the 1 Gb nic adapter as an FT logging it's currently running fine since the VM is'nt doing anything at the moment . we're planning to upgrade the nic to a 10 Gb nic i did some search yesterday and i found this one http://www.ebay.com/itm/Emulex-Lightpulse-LP21002-Dual-Port-10GB-PCI-E-CNA-fcoe-vmware-esx-/12113789...

Emulex Lightpulse LP21002 Dual Port 10GB PCI-E CNA

will this be a good nic to use with a cisco 4500 catalyst switch we need also to get some sfps to be able to comunicate with the server on a 10 Gb i found this refernce

SFP-10G-LR 10GBASE-LR Small Form-factor Pluggable Plus (SFP+).

anyone can confirm the compatibility with the cisco and the Emulex card ?

thanks