stony007_de
Enthusiast
Enthusiast

Disaster-Master-Plan with 2xEVA6000

in my Company run 10 ESX3.2 which stored their information in a EVA6000.

To prevent a disaster (loosing data or a high downtime), we purchased a second EVA. This Storage has the same configuration und the same hardware like the first.

The second SAN should be placed in another house, which is connected by FC with the first.

Five of the ten ESX Server will be placed In the "Backup-House".

My plan is:

- the EVA's are replicated 1:1 in every time

- If one of the eva is dieing(for example: one place is burn down), the second further run without downtime and present all data to the rest of the ESX Server

- the SAN stored VMs should restarted on the second place with the VM-HA-Service, with the current last state before the system crashed(exclude the fact of database or filesystem errors)

Now the question:

Has anybody use a strategy to resolve my plan?

Is that possible?

Are ther any software solutions necessary?

friendly greets

Stony

0 Kudos
15 Replies
habibalby
Hot Shot
Hot Shot

Hi,

What you can do as follows:

  1. Connecting your Primary Site with the DR Site, where your Indentical Servers, Sotrage and networking resides.

  2. Replicate the EVA Storage from the Primary Site to the DR Site

  3. http://www.vizioncore.com/vReplicator.html

Hope it helps

Best Regards, Hussain Al Sayed Consider awarding points for "correct" or "helpful".
0 Kudos
aleph0
Hot Shot
Hot Shot

the switch will not be automatic.

I've built a solution like that to a customer site (2 EMC SAN with FC between 2 site): syncronous mirror.

If one site burn (ie SAN on burned site is dead) the plan to recovery is:

manually promote the mirrored LUN to primary

manually assign the new primary LUN to hosts

in an host Configuration Tab --> rescan for LUN / VMFS

The new LUN will be found with name snap-0000000X-name_of_the_faulted_LUN

remove from inventory each inaccessible vm

(if you want) rename snap-00000X-name_of_the_faulted_LUN in name_of_the_faulted_LUN

Browse the new discovered LUN and re-register each vmx in there

start the re-registered virtual machine

with EMC I need to edit LUN.enable_resignature parameters in order to have a working DR!!!!

HTH

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
stony007_de
Enthusiast
Enthusiast

wow... do you kidding me?

there is no clean possibility to run VMs in a disaster, automaticly from a mirrored Store?

With the vReplicator i can replicate the VM to another ESX Host.

But what is happend if one of the Sites gets a crash ?

What i have to configure on the EVA that the SAN is 100% syncron?

Which HP features i have to buy (Continuous Access,etc...)

0 Kudos
aleph0
Hot Shot
Hot Shot

I'm not a SAN expert but, i think, that there will be a primary LUN and a slave synced LUN. You have to present the primary LUN to hosts in order to work fine (AFAIK) and you'll promote the synced LUN in case of disaster.

The procedure that I've written above is completely automatizable with script for the VI tasks. I don't know what you can automatize in EVA SAN (promotion of slave LUNs to primary).

The problem is what event will trigger the two automatizated tasks (VU and LUNs).

Another thing to take in mind is: "do I want, in case of disaster, that's everything is automatically triggered?"

"what about the tiers of services. am I so confident that the plan will work for every kind of disaster?"

"what if the recovery procedure is triggered by mistake?"

take in mind that the procedure that I've described and done manually in my previous post take about 5 minutes per LUN and 3 minutes per vm (less than 10 secs per vm if bash-scripted)

cheers

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
aleph0
Hot Shot
Hot Shot

what I mean is that you'll have an env like in the attached image where half of the SAN in pri site is synced in the secondary site, and half of the SAN in the secondary site is synced in primary.

In this way you'll have primary LUN(s) in the first site and primary LUN(s) in the second site each synced in the "other" site (each site is both primary and secondary).

hosts in primary site can access primary Lun(s) in secondary site and hosts in secondary site can access primary LUN(s) in primary site.

hosts in primary and secondary site are both in production!

HTH (but i'm not sure to explain well) Smiley Happy

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
aleph0
Hot Shot
Hot Shot

hi,

have you managed to draw your disaster master plan?

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
christianZ
Champion
Champion

With CA you will be able to do it like aleph0 described.

0 Kudos
aleph0
Hot Shot
Hot Shot

I've done it with ESX and nothing else ...

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
christianZ
Champion
Champion

...and with EMC mirroring features. CA (Continues Access) is the mirror feature from HP.

0 Kudos
aleph0
Hot Shot
Hot Shot

Sorry ChristianZ... I belive that was some sort of sw from Computer Associates...

And Yes, You'll need built in EMC Mirroring... I don't know Continues Access so probably you will need it with youe EVA

Cheers

HTH

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
Mork
Enthusiast
Enthusiast

I have done the same aleph0 for some VM's, but other VM's host applications that have built-in redundancy, so I need active VM's in both Production and DR sites.

I've ended up with some Datastores being replicated to the DR site and some Datastores at each site that are not replicated.

I also have a VirtualCenter DR setup configured so that the SQL DB is backed up to the local disk and I use robocopy to transfer the backups to the DR VirtualCenter server. This is a VM running live but with the VirtualCenter services stopped.

In a DR scenario, I restore the latest copy of my SQL backup and bring the VirtualCenter services online, then swap the replicated Datastores so the Primary is at the DR site, then remove all the offline VM's and add them to the inventory at the DR site.

One thing I did find though is that after you rescan at the DR site, you should rename your Datastore back to its original name (without snapxxxxxx- in front) otherwise you may need to edit the vmx files for the VM's.

Cheers, Pete

0 Kudos
aleph0
Hot Shot
Hot Shot

Yep! Smiley Happy

It's important to have a sleeping configured Virtual Center! sometimes this part is forgotten and it's a vital part of the DR master plan.

cheers

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
stony007_de
Enthusiast
Enthusiast

hmm that all, sounds like there is no professional solution for the Disaster-Task?

what it with the "VMWare Site Recovery Manager" ?

has anybody more information aubout this plugin?

i only found: -->

:smileycool: Stony

0 Kudos
aleph0
Hot Shot
Hot Shot

pls, describe what a "professional solution for disaster recovery" is?

is something that let the system do whatever the system wants while you are doing something else?

VMSRM is a DR recovery "orchestrator" and you can use it, afaik, to "design" your master plan without forgetting something (define the tier of service, what server must be restarted first, etc...)

\aleph0

____________________________

###############

If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!

\aleph0 ____________________________ http://virtualaleph.blogspot.com/ ############### If you found this information useful, please consider awarding points for "Correct" or "Helpful". Thanks!!!
0 Kudos
MZ11
Enthusiast
Enthusiast

Hi,

i'm sure stony007_de doesn't want a Disaster Recovery.

I think he wants a solution that allows the loss of one storage or location without downtime (if a VM is on a server on disaster location a restart is required).

Our customer wants a solution like this too, but the solutions from the major vendors are quiete rare.

Because the primary problem is the storage I think storage virtualization is the best solution. But I don't know how to deal with HP EVAs to do this.

mz1

0 Kudos