DR Site solution - recommendations wanted

[BjornJohansson]

Hello all,

I would really appreciate some input on our Disaster Recovery Solution from a VMware standpoint. Hopefully we get VMware SRM, but we'll have to wait for our new SAN first. Until then we have to rely to current solution. We have tried the plan below but I would like additional comments and suggestions.

Is there any obvious technical problems with this setup?

HQ

HP C7000 + BL480c Blade servers

EVA 3000

ESX 3.5 U3 (I know, not supported on EVA3000)

DR Site

HP C7000 + BL480c Blades

EVA4000

ESX 3.5 U3

VC 2.5 U3

Syncronization is made with HP Continous Access over a 2 mbit fibre line.

This is our plan activate the DR site, and fall back to HQ site.

1) Sync between sites are stopped and SAN APA is activated on DR Site

2) Present LUN "TestLUN" to DR Site ESX hosts

3) Activates LVM.EnableResignature 1 on all hosts and reboots

4) Adds Datastore called "snap-xxxxxxxxxxTestLUN" to ESX hosts

5) Browse to VM's .vmx and adds it to inventory

6) Starts VM and chose "keep indentifier"

7) Adds an text file on VM's desktop

😎 Shuts down the VM and reverse the process by activating CA sync

Well that is basically it. Any thoughts?

Thanks!

/Björn

[vSeanClark]

How many VMs do you have on that setup? If it's not that many, I think Veeam Backup (which does replication as well) might be a better solution, since it takes care of the registering of VMs for you. Just depends on your goals (RPO/RTO) for this environment and what the app requirements are.

Sean Clark - -

[christianZ]

This is our plan activate the DR site, and fall back to HQ site.

1) Sync between sites are stopped and SAN APA is activated on DR Site

2) Present LUN "TestLUN" to DR Site ESX hosts

3) Activates LVM.EnableResignature 1 on all hosts and reboots

that should be done only on one esx host when all your volumes are shared to all esx hosts - I think

4) Adds Datastore called "snap-xxxxxxxxxxTestLUN" to ESX hosts

adding will be not needed - I think - only rescanning

5) Browse to VM's .vmx and adds it to inventory

6) Starts VM and chose "keep indentifier"

7) Adds an text file on VM's desktop

😎 Shuts down the VM and reverse the process by activating CA sync

Just my thoughts.

[BjornJohansson]

Thank you guys for your input!

Can someone please confirm wheter LVM.EnableResignature 1 should be activated on one host?

During the tests I had some problems after the tests with some VM's turning unavailable. Can this be related to enabling on all hosts? Please see my old thread here: http://communities.vmware.com//thread/177263?tstart=0

vSeanClark, we currently have two clusters, one on each site. We have about 40 VM's on HQ. We have physical, SAN-attached servers that are critical. Therefore we use a storage level replication. Probably more expensive but opens up for a more enterprise class solultion like SRM.

[malaysiavm]

I will suggest you to reconsider the need of storage replication some times, as many of the environment we study, the DR solution from Vendor must not necessary cost and effective, depend what DR solution we want to deploy. On my experience, my DR environment is virtualized and the production server could be on physical or Virtual and we slowly consider the best option as we found that the application or databases we use is capable for certain technology to be replicated and synchronized in DR without require some heavy investment to our environment especially on WAN and SAN.

Just my 2 cents to share here.

Malaysia VMware Communities -

[Texiwill]

Hello,

LunResignature is only necessary if you are presenting a Mirror of a LUN from the SAME SAN to the ESX hosts. Now Continuous Access may make both SANs appear as one but they really are not. It depends if the LUN gets marked as a Mirror or not. If it is marked as a mirror then you need LunResignature enabled on conversion from a mirror LUN back to a regular LUN.

Best regards,
Edward L. Haletky
VMware Communities User Moderator, VMware vExpert 2009
====
Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.
Blue Gears and SearchVMware Pro Blogs -- Top Virtualization Security Links -- Virtualization Security Round Table Podcast

[vSeanClark]

I like 's thought here and would like to add another $.02. A hybrid approach to replicating key services might be best. You have have 3 methods of VM replication at your disposal (roughly): SAN replication, ESX host replication (Veeam or Vizioncore), and Application level replication (DFS, Exchange CCR, SQL Server 2005 replication). I would also like to throw one other option out, backing up full servers to a DataDomain box and replicating those backups over the WAN.

To Malaysiavm's point, if you have DFS replication, Exchange replication and SQL replication all setup and running. You might have just covered all your mission critical assets without having to pay for expensive SAN replication licenses, AND you will have implemented a replication scheme that is fully application aware. Many times your SAN replication processes does not guarantee application aware snapshots, so you recovery site will just be crash consistent. Unless you have a special storage configuration for your VMs and have the right software to coordinate with VMware AND the app before taking SAN snapshots for replication.

SRM is definitely cool and the Cadillac solution, ESPECIALLY if your SAN's snapshot and replication features support your business critical applications.

But, it's good to know all your options for meeting your RTOs and RPOs, especially in a down economy with budgets getting slashed.

Sean Clark - -

[christianZ]

...in addition one question to VC - have you only one VC server placed on DR and managing both sites?

[malaysiavm]

I am absolutely agree with the point from Sean here

Malaysia VMware Communities -

[vSeanClark]

I would leave vCenter server at the primary site so that it performs the best for your daily use. Yould could also use it to manage your DR ESX hosts. You would then want replicate vCenter to the DR site. I prefer vCenter as a VM for this reason.

If you go with SRM, then you will have to get a second vCenter server - one for protected side and one for recovery site.

I hope that is helpful. -Sean Clark

[BjornJohansson]

Hello,

LunResignature is only necessary if you are presenting a Mirror of a LUN from the SAME SAN to the ESX hosts. Now Continuous Access may make both SANs appear as one but they really are not. It depends if the LUN gets marked as a Mirror or not. If it is marked as a mirror then you need LunResignature enabled on conversion from a mirror LUN back to a regular LUN.

</a>

Hello Texiwill and thank you for your input! I think it's is considered a mirror. As far as I've understood when a LUN is replicated it has the same signature as the source. And therefore the ESX (or ESX's?) on our recovery site needs to write a new signature in order to access the VMFS on it. I did try to not to use the resigning setting but without luck. The problem is the previously mentioned problem with invalid VM's which I strongly suspect to be connected with this setting. So, do I need to do it on all ESX's or just one? Thx!

Also I would like to thank all that contributed with input! Even though we made the decision with SAN level replication your input is relevant for the discussion!

Thanks!

/Björn

[BUGCHK]

You are replicating from an EVA3000 (HSV100/HSV101) to an EVA4000 (HSV200). Both arrays return different SCSI inquiry strings which VMware ESX does care about so you will have to resignature, or, provided that you do NOT present any snapshots, set the LVM.DisallowSnapshotLUN parameter as a workaround.

[aleph0]

Hi Björn,

see what I'm doing while waiting for SRM plugin for IBM DS8300. in this post

Anyone that want to give other suggestions/feedback on procedure I'm using is appreciated.

cheers

\aleph0

____________________________

(in italian)

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

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