VMware Cloud Community
JasonVmware
Enthusiast
Enthusiast
Jump to solution

Question About Multiple Site Design

Hello All,

We currently have 2 Equallgoic SAN's one in the main office and one in are sub office. We have 3 ESX hosts at each location for a total of 6 ESX hosts. All VM's are stored on the Equallogic SANS and all hosts will tie into a Virtual Center Server at the head office. Each site has a few VM's that run for that office and replicate those VM's to the other office for disaster recovery purposes. I have attached a ruff diagram of how it is setup.

Now for the questions:

1.) With the VM's being spread across two sites equally how will Virtual Center handle High Availablilty across the VPN?

2.) If HA does trigger due to a host failure at one of the site how will it decide which server the VM's will start up on, and if it decides to start up on another host in another site how will it know to use the replication data

3.) We currently don't have site recovery manager so the recovery is a manual process. What would be the best manual restore process?

Reply
0 Kudos
1 Solution

Accepted Solutions
RussellCorey
Hot Shot
Hot Shot
Jump to solution

1.) With the VM's being spread across two sites equally how will Virtual Center handle High Availablilty across the VPN?

HA will work but only within a given site. i.e. if a VM is at site A and a host in site A fails, it will only power up at site A. If you want it to go to site B you should look into SRM.

*2.) If HA does trigger due to a host failure at one of the site how

will it decide which server the VM's will start up on, and if it

decides to start up on another host in another site how will it know to

use the replication data*

It won't try to start up on the other side and use the replicated data because the replication target is most likely read only.

**<span class="jive-thread-reply-body-container">3.) We currently don't have site recovery manager so the recovery is a*

manual process. What would be the best manual restore process?*

A complete site failure is going to look something like this:

1. Break replication

2. Make your replicated LUNs at your recovery site read/write capable

3. Present replicated LUNs to your ESX server hosts

4. Configure ESX server host to resignature LUNs and rescan or add datastores and resig (depending on ESX3 or 4)

5. Browse datastore or use vmware-cmd to register virtual machines on the replicated LUNs

6. Do a quick sanity check (make sure vSwitch/portgroup configuration is acceptable)

7. Power VMs on 1 at a time. Since its an IPSEC connection I'm doubting you've spanned VLANs so you're going to have to re-IP everything as it comes up

8. Change DNS settings

9. Adjust upstream/downstream dependencies as required.

SRM will automate everything up to 8 and can typically be scriped to handle 9. This isn't 100% comprehensive but should give you an idea.

edit to add your next questions:

*should the LUNS from

both sites be shared to all ESX servers? or would it be better to just

have the local ESX servers have access to its local LUNS ?*

Your local ESX servers should have access to only local storage devices unless you have a very low latency and fast connection between sites (say &lt;10ms and at least gigabit) as well as having that connection be a layer 2 connection. That way your VLANs/subnets/etc are still the same on the other side.

View solution in original post

Reply
0 Kudos
3 Replies
JasonVmware
Enthusiast
Enthusiast
Jump to solution

Does anyone know how the HA would work? and should the LUNS from both sites be shared to all ESX servers? or would it be better to just have the local ESX servers have access to its local LUNS ?

Reply
0 Kudos
RussellCorey
Hot Shot
Hot Shot
Jump to solution

1.) With the VM's being spread across two sites equally how will Virtual Center handle High Availablilty across the VPN?

HA will work but only within a given site. i.e. if a VM is at site A and a host in site A fails, it will only power up at site A. If you want it to go to site B you should look into SRM.

*2.) If HA does trigger due to a host failure at one of the site how

will it decide which server the VM's will start up on, and if it

decides to start up on another host in another site how will it know to

use the replication data*

It won't try to start up on the other side and use the replicated data because the replication target is most likely read only.

**<span class="jive-thread-reply-body-container">3.) We currently don't have site recovery manager so the recovery is a*

manual process. What would be the best manual restore process?*

A complete site failure is going to look something like this:

1. Break replication

2. Make your replicated LUNs at your recovery site read/write capable

3. Present replicated LUNs to your ESX server hosts

4. Configure ESX server host to resignature LUNs and rescan or add datastores and resig (depending on ESX3 or 4)

5. Browse datastore or use vmware-cmd to register virtual machines on the replicated LUNs

6. Do a quick sanity check (make sure vSwitch/portgroup configuration is acceptable)

7. Power VMs on 1 at a time. Since its an IPSEC connection I'm doubting you've spanned VLANs so you're going to have to re-IP everything as it comes up

8. Change DNS settings

9. Adjust upstream/downstream dependencies as required.

SRM will automate everything up to 8 and can typically be scriped to handle 9. This isn't 100% comprehensive but should give you an idea.

edit to add your next questions:

*should the LUNS from

both sites be shared to all ESX servers? or would it be better to just

have the local ESX servers have access to its local LUNS ?*

Your local ESX servers should have access to only local storage devices unless you have a very low latency and fast connection between sites (say &lt;10ms and at least gigabit) as well as having that connection be a layer 2 connection. That way your VLANs/subnets/etc are still the same on the other side.

Reply
0 Kudos
mreferre
Champion
Champion
Jump to solution

Jason,

I am working on a document that is supposed to cover all this. Not so much from a product perspective but from an architectural perspective (although I mention some of the San/Storage products out there that are capable of implementing specific features). I don't cover Equallogic from a product perspective but if you (or your vendor) can map what it can and cannot do you should have your answers.

The document covers specifically two scenarios: Campus (which would be your scenario) and Globe/DR. I have included both VMware and Hyper-V per each of the scenario.

I want to post it on my blig (www.it20.info) but it won't happen any time soon (hope in a few weeks). The doc is far from being complete but I have a draft of the Campus section ready. If you are interested just PM me and I will send you an alpha draft for review (I'd like to understand if I am on the right track and if the content can be beneficial for the community).

Massimo.

Massimo Re Ferre' VMware vCloud Architect twitter.com/mreferre www.it20.info
Reply
0 Kudos