Discussions specific to the VMware vSphere hypervisor
CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 06, 2013 5:51 pm

Hi,

I am starting a new project to backup and replicate our vSphere environment using B&R v7. I wondered if I could run my proposed setup by you all to get any feedback on improvements/alterations I could make or any issues that I may have and need to address. We don't use B&R in production yet but use another product for vSphere to tape backups. Sorry for the length of this but thought it best to give you all the details up front and thanks in advance for any help provided.


Current/Production Environment:
-------------------------------

- vSphere 5.0 Enterprise (upgrading to 5.1 soon) with 6 ESXi hosts on primary site
- vCenter on Windows 2008 R2 Server VM using SQL for DB
- Dell iSCSI 10GBE SAN storage, around 30TB capacity
- Approx 125 Windows VMs some thin, some thick, B&R tells me a total of 18.5TB (will be increasing when we virtualise more soon)
- Windows 2003/2008 R2 mixed mode domain (not replicating any primary site DCs)
- Applications include 3rd Party, File Data, Exchange 2010 DAG (2 VMs) not using RDMs, SharePoint, Ecommerce (IIS), SQL 2005/2008, etc
- We also host many websites, in a multiple DMZ environment protected by firewalls (separated by vmnic's/vswitches in the vSphere environment)
- Servers all use DHCP reservations, IP range 172.18.x.x, approx 20 server VLANs
- Physical Windows 2008 R2 Backup Server with VMFS Datastores directly presented to it for SAN mode backups
- 100Mbit Internet Link
- Primary Site Backup Storage - Dell iSCSI 10GBE SAN - 23 x 3TB SATA drives in RAID 6 (1 x Hot Spare) around 57TB usable. Connected to Backup Server as NTFS volume


DR Site:
--------

- vSphere 5.1 Essentials Plus with 1 ESXi host
- vCenter on Windows 2008 R2 (or 2012) Server VM using SQL for DB
- Deploy Veeam Backup Proxy on vCenter or other VM
- 50Mbit Internet Link (will be connected to primary site by site to site VPN using Cisco firewalls)
- DR Storage - Dell iSCSI 10GBE SAN - 23 x 3TB SATA drives in RAID 6 (1 x Hot Spare) around 57TB usable
- Will be creating a DR Active Directory Site with a VM DC connected to the production domain that will also run DHCP, IP range 172.20.x.x


Business Requirements:
----------------------

1. There is no definite RPO or RTO, but at the moment we do nightly tape backups so this will be a major improvement on that.

2. At least twice a year some members of the business want to visit the DR site (where there are PCs/desks etc they can use) and test all applications and services without affecting the production replication. After the test any changes they have made can be lost (it's only a test). Unfortunately the automated tests that B&R can do are not enough for internal audit, they want users to actually test applications and services at the DR site.


My Proposal:
------------

1. Backup to Primary Site Backup Storage (NTFS) every 4 hours approx, stored as Veeam backup files.

2. Replicate from Primary Site Backup Storage (as seed) to DR Storage (VMFS Datastore) every 4 hours approx (after previous job I guess). I would prefer this to be replication I think so the servers actually exist at the DR site as VMs (VMDKs). I suppose this a debatable point because I can't use WAN acceleration if I use replication at present.

3. Re-create the Primary Site Network VLANs, DHCP scopes etc, but with 172.20.x.x range instead of 172.18.x.x. Configure DR site DHCP server identically to Primary Site but with different IP range to avoid conflicts etc.

4. Use SureReplica for business requirement 2, for when users come to test, which would be separated from production replicated VMs. One problem with this is giving users access to the SureReplica VMs through the proxy and also I would need a DHCP server in that Virtual Lab otherwise none of these VMs will get IPs.


Questions/Discussion:
---------------------

I have already setup the primary site backup storage, created a full backup and some incremental backups (all ok so far) and also on the same site connected the DR storage to a test ESXi host and setup a replication job to replicate from the primary site backup storage (as seed) to the DR storage. What I didn't expect is that it is using the seed backup repository but also then connecting to the production VMs to get the latest snapshot and also replicating that, but not updating the repository with that latest backup. I'm not certain I'm doing this the right way so really hoping someone is going to tell me whether I am or whether I should be doing it slightly differently? For info, that first full backup took that 18.5TB to around 6.8TB and took 22 hours.

The replication is seems slow, maybe it's because it's the first full run, not sure? I've only connected the DR storage to my test ESXi host by one 1GBit connection (and the test ESXi host is connected to the LAN by just one 1Gbit connection), I thought that around 100MB/s transfer would be ok and of course I didn't expect it to be snapshoting the production VMs as part of the replication. I also thought that when it's in the DR site it won't be getting anywhere near 100MB/s but actually it's currently running at about 40-50MB/s, still way in excess of the ~5MB/s bandwidth I will have to the DR site when the storage is installed and everything configured on the DR site. Is this what you would expect? Anyway to stop it taking the snapshot from the production VM when replicating?

I suppose there is many ways I could do this all which may not be right or wrong. I just thought that maybe I should divide up the storage at DR into VMFS and NTFS and use the WAN accelerated backup to the NTFS storage and then replicate the VMs from that into actual VMs for testing/failover etc? Not how I would prefer I think, especially as it will use more storage all the time and I won't be able to show management all our replica VMs sitting there on the DR vSphere environment just waiting to be used.

Would point 4 of my proposal above work?

Any thoughts?

Thanks for reading all of this and if I have missed anything or if anything isn't clear let me know.

Simon

Vitaliy S.
Product Manager
Posts: 22289
Liked: 1413 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by Vitaliy S. » Sep 07, 2013 5:11 pm 1 person likes this post

Hello Simon,
CaptainFred wrote: Unfortunately the automated tests that B&R can do are not enough for internal audit, they want users to actually test applications and services at the DR site.
If users can automate application verification via scrips, then you can use these scripts as a part of the SureBackup job as well.
CaptainFred wrote:What I didn't expect is that it is using the seed backup repository but also then connecting to the production VMs to get the latest snapshot and also replicating that, but not updating the repository with that latest backup. I'm not certain I'm doing this the right way so really hoping someone is going to tell me whether I am or whether I should be doing it slightly differently? For info, that first full backup took that 18.5TB to around 6.8TB and took 22 hours.
But why it should update the backup repository you were using as a seed with the latest VM data? Seeding is just a one time operation that basically restores the VM from the backup and then maps production VM to the restored one to update it with the recent VM changes. This operation should be used when you do not want to transfer the entire VM over the WAN link. Once initial job pass is done, you do not need this seed anymore.
CaptainFred wrote:The replication is seems slow, maybe it's because it's the first full run, not sure? I've only connected the DR storage to my test ESXi host by one 1GBit connection (and the test ESXi host is connected to the LAN by just one 1Gbit connection), I thought that around 100MB/s transfer would be ok and of course I didn't expect it to be snapshoting the production VMs as part of the replication. I also thought that when it's in the DR site it won't be getting anywhere near 100MB/s but actually it's currently running at about 40-50MB/s, still way in excess of the ~5MB/s bandwidth I will have to the DR site when the storage is installed and everything configured on the DR site. Is this what you would expect? Anyway to stop it taking the snapshot from the production VM when replicating?
After you do replica mapping/seeding, Veeam replication jobs has to calculate all the VM digests to detect what blocks have changed after seeding. Can you please tell me where you store replica metadata? Is it on the source site repository? BTW, can you confirm that you're using proxy servers on both sites?
CaptainFred wrote:...and of course I didn't expect it to be snapshoting the production VMs as part of the replication.... Anyway to stop it taking the snapshot from the production VM when replicating?
Not sure I understand the question... If you want to backup/replicate a VM you need to create a snapshot first and then start transferring VM data.
CaptainFred wrote:I just thought that maybe I should divide up the storage at DR into VMFS and NTFS and use the WAN accelerated backup to the NTFS storage and then replicate the VMs from that into actual VMs for testing/failover etc? Not how I would prefer I think, especially as it will use more storage all the time and I won't be able to show management all our replica VMs sitting there on the DR vSphere environment just waiting to be used.

Would point 4 of my proposal above work?
Well...certainly using backup copy jobs with WAN acceleration enabled to transfer your backups to the offsite location is a recommended way, but I don't completely get what do you mean by "replicate the VMs from that into actual VMs for testing/failover etc". Do you want to run a backup restore job from these files? If this is the case, then using replication jobs is definitely a preferable way to have "ready to run" and verified copies of production VMs on the DR site.

Hope this helps!

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 09, 2013 1:48 pm

Thanks for your reply.
CaptainFred wrote: Unfortunately the automated tests that B&R can do are not enough for internal audit, they want users to actually test applications and services at the DR site.
Vitaliy S. wrote:If users can automate application verification via scrips, then you can use these scripts as a part of the SureBackup job as well.
Yes I know but that’s what they want to do. They also have to test things like phones and business processes which would be slightly different because there is less people who would be utilised in the DR situation. So would my plan of using the Virtual Lab (SureReplica) work (remember I need to run the test and keep production replication ongoing)?
CaptainFred wrote:What I didn't expect is that it is using the seed backup repository but also then connecting to the production VMs to get the latest snapshot and also replicating that, but not updating the repository with that latest backup. I'm not certain I'm doing this the right way so really hoping someone is going to tell me whether I am or whether I should be doing it slightly differently? For info, that first full backup took that 18.5TB to around 6.8TB and took 22 hours.
Vitaliy S. wrote:But why it should update the backup repository you were using as a seed with the latest VM data? Seeding is just a one time operation that basically restores the VM from the backup and then maps production VM to the restored one to update it with the recent VM changes. This operation should be used when you do not want to transfer the entire VM over the WAN link. Once initial job pass is done, you do not need this seed anymore.
I see. I think I have misunderstood the point of seeding. I thought the point was so that only one backup is required to be run from the production storage, so it’s not impacted by 2 jobs.

Can I ask how you would recommend I deploy in my situation? Am I actually going to need to have 2 jobs running every 4 hours (one backup job one replication job)? Will they conflict if they happen to get to the same VM at the same point in time?
CaptainFred wrote:The replication is seems slow, maybe it's because it's the first full run, not sure? I've only connected the DR storage to my test ESXi host by one 1GBit connection (and the test ESXi host is connected to the LAN by just one 1Gbit connection), I thought that around 100MB/s transfer would be ok and of course I didn't expect it to be snapshoting the production VMs as part of the replication. I also thought that when it's in the DR site it won't be getting anywhere near 100MB/s but actually it's currently running at about 40-50MB/s, still way in excess of the ~5MB/s bandwidth I will have to the DR site when the storage is installed and everything configured on the DR site. Is this what you would expect? Anyway to stop it taking the snapshot from the production VM when replicating?
Vitaliy S. wrote:After you do replica mapping/seeding, Veeam replication jobs has to calculate all the VM digests to detect what blocks have changed after seeding. Can you please tell me where you store replica metadata? Is it on the source site repository? BTW, can you confirm that you're using proxy servers on both sites?
Yes the “Repository for replica metadata” is in the source site backup repository. Is this correct?

I haven’t deployed the DR proxy server yet because I’m preparing the replica VMs on the DR storage locally before I physically take it to the DR site. There is a 1Gbit connection between all the components at the moment, it should easily be able to utilise 100% of a single 1Gbit connection.
CaptainFred wrote:...and of course I didn't expect it to be snapshoting the production VMs as part of the replication.... Anyway to stop it taking the snapshot from the production VM when replicating?
Vitaliy S. wrote:Not sure I understand the question... If you want to backup/replicate a VM you need to create a snapshot first and then start transferring VM data.
This is kind of related to my misunderstanding of seeding, I think, see above.
CaptainFred wrote:I just thought that maybe I should divide up the storage at DR into VMFS and NTFS and use the WAN accelerated backup to the NTFS storage and then replicate the VMs from that into actual VMs for testing/failover etc? Not how I would prefer I think, especially as it will use more storage all the time and I won't be able to show management all our replica VMs sitting there on the DR vSphere environment just waiting to be used.
Would point 4 of my proposal above work?
Vitaliy S. wrote:Well...certainly using backup copy jobs with WAN acceleration enabled to transfer your backups to the offsite location is a recommended way, but I don't completely get what do you mean by "replicate the VMs from that into actual VMs for testing/failover etc". Do you want to run a backup restore job from these files? If this is the case, then using replication jobs is definitely a preferable way to have "ready to run" and verified copies of production VMs on the DR site.
Ok I am correct in saying that this is the recommended approach?:

Primary Site contains Production Storage and Production running VMs
Use Backup Job – to backup VMs to Backup Repository (NTFS volume) located on Primary Site
Use Replication Job – to Replicate VMs to ESXi/vSphere (VMFS datastore) located on DR Site


Extra questions:

A. Can the Production and DR Site Servers that are running B&R and/or Proxies be physical or virtual or a combination? Eg. Physical on Production Site, Virtual on DR Site?

B. Can the Production and DR Site Servers that are running B&R and/or Proxies be running on different operating systems?

C. When I select a VM and the choose the option to “Restore guest files (windows)” I can restore files and if it’s an Exchange server I can restore email out of databases, but if it’s SQL server the option to restore database(s) doesn’t appear, do I have to do something different if the server contains SQL? Note I have not selected “Enable application aware image processing” or “Enable guest file system indexing” because it says it’s optional - do I need to enable one of these for SQL?

Thanks

Vitaliy S.
Product Manager
Posts: 22289
Liked: 1413 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by Vitaliy S. » Sep 09, 2013 3:10 pm

CaptainFred wrote:So would my plan of using the Virtual Lab (SureReplica) work (remember I need to run the test and keep production replication ongoing)?
Yes, that would definitely work.
CaptainFred wrote:Can I ask how you would recommend I deploy in my situation? Am I actually going to need to have 2 jobs running every 4 hours (one backup job one replication job)? Will they conflict if they happen to get to the same VM at the same point in time?
No issues with two jobs processing the same VM. Please read this topic for further information > Double backup
CaptainFred wrote:Yes the “Repository for replica metadata” is in the source site backup repository. Is this correct?
Yes.
CaptainFred wrote:I haven’t deployed the DR proxy server yet because I’m preparing the replica VMs on the DR storage locally before I physically take it to the DR site. There is a 1Gbit connection between all the components at the moment, it should easily be able to utilise 100% of a single 1Gbit connection.
Depends on how your network is already utilized, but yes, it should take the entire bandwidth. If not, then please look through this thread to make your jobs use the entire bandwidth > VB replication not using the entire WAN bandwidth

When you move your replicas offiste, make sure you deploy the secondary proxy server. This is essential when you're replication over the WAN link.
CaptainFred wrote:Ok I am correct in saying that this is the recommended approach?:

Primary Site contains Production Storage and Production running VMs
Use Backup Job – to backup VMs to Backup Repository (NTFS volume) located on Primary Site
Use Replication Job – to Replicate VMs to ESXi/vSphere (VMFS datastore) located on DR Site
Yes, looks good to me.
CaptainFred wrote:Extra questions:

A. Can the Production and DR Site Servers that are running B&R and/or Proxies be physical or virtual or a combination? Eg. Physical on Production Site, Virtual on DR Site?

B. Can the Production and DR Site Servers that are running B&R and/or Proxies be running on different operating systems?

C. When I select a VM and the choose the option to “Restore guest files (windows)” I can restore files and if it’s an Exchange server I can restore email out of databases, but if it’s SQL server the option to restore database(s) doesn’t appear, do I have to do something different if the server contains SQL? Note I have not selected “Enable application aware image processing” or “Enable guest file system indexing” because it says it’s optional - do I need to enable one of these for SQL?
A. Yes, no limitations with that, but I would prefer to have virtual proxy server to leverage hotadd capability during restore and backup process.
B. Yes.
C. To restore SQL Server items, you should be using SQL AIR wizard > VMware : [FAQ] FREQUENTLY ASKED QUESTIONS

Also I would strongly recommend using application-aware image processing for all Windows VMs, see this topic for more details > >>> READ FIRST : [FAQ] FREQUENTLY ASKED QUESTIONS <<<

Thanks!

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 09, 2013 6:18 pm

Vitaliy S. wrote: Yes, that would definitely work.
Ok so how do I deal with access to the Virtual Lab for the DR test users without causing conflicts bearing in mind that all the servers are DHCP and that DHCP runs on the DC which I'm not replicating and not only that the servers in the Virtual Lab will need a domain to connect to so somehow I will need a DC in there as well? Sounds very complicated - I don't see how it will work. Am I explaining this properly? :?
CaptainFred wrote:Can I ask how you would recommend I deploy in my situation? Am I actually going to need to have 2 jobs running every 4 hours (one backup job one replication job)? Will they conflict if they happen to get to the same VM at the same point in time?
Vitaliy S. wrote:No issues with two jobs processing the same VM. Please read this topic for further information > Double backup
According to that thread, they would conflict if the replication job happened to get to the same VM as the backup job was working with at the time? Specifying different times won't necessarily help because both jobs will be working through a list of over 100 VMs. This is why I thought you could backup to local backup repository and then replicate from that offsite so that only ONE job is running on the production storage and ONE backup per VM not two.

CaptainFred wrote:Extra questions:
A. Can the Production and DR Site Servers that are running B&R and/or Proxies be physical or virtual or a combination? Eg. Physical on Production Site, Virtual on DR Site?
B. Can the Production and DR Site Servers that are running B&R and/or Proxies be running on different operating systems?
C. When I select a VM and the choose the option to “Restore guest files (windows)” I can restore files and if it’s an Exchange server I can restore email out of databases, but if it’s SQL server the option to restore database(s) doesn’t appear, do I have to do something different if the server contains SQL? Note I have not selected “Enable application aware image processing” or “Enable guest file system indexing” because it says it’s optional - do I need to enable one of these for SQL?
Vitaliy S. wrote:A. Yes, no limitations with that, but I would prefer to have virtual proxy server to leverage hotadd capability during restore and backup process.
B. Yes.
C. To restore SQL Server items, you should be using SQL AIR wizard > VMware : [FAQ] FREQUENTLY ASKED QUESTIONS

Also I would strongly recommend using application-aware image processing for all Windows VMs, see this topic for more details > >>> READ FIRST : [FAQ] FREQUENTLY ASKED QUESTIONS <<<

Thanks!
A. Not sure I understand - better to have virtual proxy server on the production site or DR site? If my physical backup server on the production site can perform SAN mode backups surely this is better than hotadd? Or is it not?


New Question:

D. When I select a datastore or cluster to backup, it builds the list of VMs to backup and goes through them in what seems to be a random order. How can I get it to go through them alphabetically (or even another order I specify)?

Thanks for your help :)

Vitaliy S.
Product Manager
Posts: 22289
Liked: 1413 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by Vitaliy S. » Sep 09, 2013 9:06 pm

CaptainFred wrote:Ok so how do I deal with access to the Virtual Lab for the DR test users without causing conflicts bearing in mind that all the servers are DHCP and that DHCP runs on the DC which I'm not replicating and not only that the servers in the Virtual Lab will need a domain to connect to so somehow I will need a DC in there as well? Sounds very complicated - I don't see how it will work. Am I explaining this properly? :?
If you're going to verify an Exchange server, then of course you will need a DC and a DNS VMs in the application group up and running prior starting the Exchange verification job. Basically, you need to have all VMs replicated that verified application depends on.

Here is more info on the network routing, please look it through > VMware : [FAQ] FREQUENTLY ASKED QUESTIONS

Also if you take a look at our User Guide (page 268) you will see how network is configured for virtual labs.
CaptainFred wrote:According to that thread, they would conflict if the replication job happened to get to the same VM as the backup job was working with at the time? Specifying different times won't necessarily help because both jobs will be working through a list of over 100 VMs. This is why I thought you could backup to local backup repository and then replicate from that offsite so that only ONE job is running on the production storage and ONE backup per VM not two.
Yes, there will be a conflict, but no jobs will fail. The VM will stand in a queue in the second job. Once first job is complete, your second job will start processing the selected VM, so there is no need to bother with job scheduling.
CaptainFred wrote:A. Not sure I understand - better to have virtual proxy server on the production site or DR site? If my physical backup server on the production site can perform SAN mode backups surely this is better than hotadd? Or is it not?
If you have a proxy server acting in the SAN mode on the main site, then of course this is preferable. I was referring to the DR site, where hotadd proxy can be used to write VM data to the datastore directly.
CaptainFred wrote:D. When I select a datastore or cluster to backup, it builds the list of VMs to backup and goes through them in what seems to be a random order. How can I get it to go through them alphabetically (or even another order I specify)?
Do you want them to be processed or displayed in the alphabetical order? Btw are you adding VMs explicitly or in the container? The recommended way to add VMs to the job is to use VM folders/datastores/hosts etc. In this case all new VMs added to these folders (to your VI) will be picked up by the backup job automatically.

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 10, 2013 12:05 pm

Vitaliy S. wrote: If you're going to verify an Exchange server, then of course you will need a DC and a DNS VMs in the application group up and running prior starting the Exchange verification job. Basically, you need to have all VMs replicated that verified application depends on.

Here is more info on the network routing, please look it through > VMware : [FAQ] FREQUENTLY ASKED QUESTIONS

Also if you take a look at our User Guide (page 268) you will see how network is configured for virtual labs.
Ah ok that throws a spanner in my plans! I thought (possibly incorrectly) that it can cause problems snapshoting a domain controller and it’s not supported by MS. Also lots of issues can occur like USN rollback etc - http://support.microsoft.com/kb/875495

I have been reading the manual and FAQs but tbh I find this area quite complicated.
Vitaliy S. wrote: Yes, there will be a conflict, but no jobs will fail. The VM will stand in a queue in the second job. Once first job is complete, your second job will start processing the selected VM, so there is no need to bother with job scheduling.


Ok so it won’t error? It will simply wait until the VM is available?
Vitaliy S. wrote: If you have a proxy server acting in the SAN mode on the main site, then of course this is preferable. I was referring to the DR site, where hotadd proxy can be used to write VM data to the datastore directly.
Yes, I’m planning this:

Production Site – Physical Backup Server and Replication Source Proxy – Windows Server 2008 R2 SP1 - SAN mode connectivity (direct access to vSphere datastores)
DR Site – Virtual Backup Server and Replication Destination Proxy – Windows Server 2012

1. Is this the best design?

2. I could give the DR Site VM direct access to the SAN vSphere datastores or wouldn’t this work in the same way as it does for a physical server or would it make no difference in terms of performance compared to hotadd?

3. Referring to the “hotadd proxy” and the explanation on page 28 of the manual – does this method actually attach the VMDK of a VM to the Backup VM so it can copy the data straight out of the VMDK? Even though it’s still attached to the real VM?

4. Even so, the data still has to be transferred across the network so is this why it’s not as good as a physical server with direct access to the SAN/vSphere datastores?
Vitaliy S. wrote: Do you want them to be processed or displayed in the alphabetical order? Btw are you adding VMs explicitly or in the container? The recommended way to add VMs to the job is to use VM folders/datastores/hosts etc. In this case all new VMs added to these folders (to your VI) will be picked up by the backup job automatically.
I would like them to be processed in alphabetical order. I have added the cluster. Same thing happens when I select folders/datastores/etc. Is this expected or a bug?


F. I know it’s only just been released but I thought I might as well ask – can I use Windows Server 2012 R2 as the VM on the DR site?


Thanks again for your help, much appreciated.

Vitaliy S.
Product Manager
Posts: 22289
Liked: 1413 times
Joined: Mar 30, 2009 9:13 am
Full Name: Vitaliy Safarov
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by Vitaliy S. » Sep 10, 2013 1:40 pm

CaptainFred wrote:Ah ok that throws a spanner in my plans! I thought (possibly incorrectly) that it can cause problems snapshoting a domain controller and it’s not supported by MS.
There are no problems with snapshotting DC VMs, just make sure you use application-aware image processing option. Here is an existing topic for additional details > AD Controller System State Issue
CaptainFred wrote:Ok so it won’t error? It will simply wait until the VM is available
Correct.
CaptainFred wrote: Production Site – Physical Backup Server and Replication Source Proxy – Windows Server 2008 R2 SP1 - SAN mode connectivity (direct access to vSphere datastores)
DR Site – Virtual Backup Server and Replication Destination Proxy – Windows Server 2012

1. Is this the best design?

2. I could give the DR Site VM direct access to the SAN vSphere datastores or wouldn’t this work in the same way as it does for a physical server or would it make no difference in terms of performance compared to hotadd?

3. Referring to the “hotadd proxy” and the explanation on page 28 of the manual – does this method actually attach the VMDK of a VM to the Backup VM so it can copy the data straight out of the VMDK? Even though it’s still attached to the real VM?

4. Even so, the data still has to be transferred across the network so is this why it’s not as good as a physical server with direct access to the SAN/vSphere datastores?
1. Yes, you may want to additionally deploy virtual proxy servers on the main site for the redundancy reasons.
2. You can not use SAN mode to write/restore VM data to the datastore, SAN mode can only be used to retrieve VMs, so there is no sense in configuring direct SAN mode on your DR site.
3. Correct. The reason why you can attach virtual disk to another VM is a VM snapshot, it "unlocks" virtual disks, so Veeam job can back these disks up.
4. Replication jobs transfer VM data over the network, that's correct and that is why you need to have proxy servers deployed on both sites to minimize the amount of traffic sent over the WAN link.
CaptainFred wrote:I would like them to be processed in alphabetical order. I have added the cluster. Same thing happens when I select folders/datastores/etc. Is this expected or a bug?
Yes, this is expected, VMs are processed in the order returned by vCenter Server. Can you please elaborate on why do you want to process VMs in alphabetical order?
CaptainFred wrote:F. I know it’s only just been released but I thought I might as well ask – can I use Windows Server 2012 R2 ?
Please see this topic for the answer > Hyper-V 2012 R2 support

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 17, 2013 12:46 am

Ok when you say no problems snapshotting DC VMs, does this include Windows 2003 / 2008 / 2008 R2 DCs? We have a mixed mode 2003/2008 R2 domain at the present because we still have some 2003 DCs (we will be removing them eventually, hopefully soon ish).

Is it recommended then to use the application-aware image processing option but not so necessary to use the guest file system indexing option? I don't really understand the point of the indexing option if it's optional and it indexes the VM anyway. I'm guessing that enabling the application-aware image processing option will slow the backups down a bit because of the extra processing?

I suppose there is no definite reason why I would like to process VMs in alphabetical order (or a specified order) but it seems more logical to me. Is there no way around this then? Not sure why vCenter returns them in such an odd order, it's not even the order they were created... at least I don't think it is!

Thanks again for your help, sorry for the delay in getting back to this thread.

foggy
Veeam Software
Posts: 17712
Liked: 1481 times
Joined: Jul 11, 2011 10:22 am
Full Name: Alexander Fogelson
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by foggy » Sep 17, 2013 7:19 am

CaptainFred wrote:Ok when you say no problems snapshotting DC VMs, does this include Windows 2003 / 2008 / 2008 R2 DCs?
Yes, all the mentioned versions are supported. There is another huge existing topic regarding domain controller recovery, worth considering for further reading.
CaptainFred wrote:Is it recommended then to use the application-aware image processing option but not so necessary to use the guest file system indexing option? I don't really understand the point of the indexing option if it's optional and it indexes the VM anyway.
Guest OS indexing is required if you need to search for guest OS files in the selected VM backup at a specific restore point using Veeam Backup Enterprise Manager (1-click FLR functionality).
CaptainFred wrote: I'm guessing that enabling the application-aware image processing option will slow the backups down a bit because of the extra processing?
There is some processing overhead required to perfrom proper application quiescence, however this will allow for automatic DC recovery without any manual intervention or additional AD administration tasks at all (as described in the topic referred above).

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 24, 2013 11:26 am

I've been working through this further and come across a few other issues/questions:

1) Performing Storage vMotion of a VM resets CBT so it backs up the whole VM again. Is this true? Seems like a big problem because we occasionally move VMs about to optimise storage and performance, some of which are several TBs so then it would have to try and send several TBs across the WAN link? This is bad enough if it was just the backup but sending that amount of data even with compression and dedupe across our 50Mbit WAN link is not going to happen! Maybe I'm missing something here. I realise this is more a VMware issue.

2) When I setup my replication job and use a backup as a replica seed is this meant to be just a one off when first setting it up to help not copy as much data across the slow WAN link or a permanent going forward solution?

3) Regarding the design, I saw this: http://helpcenter.veeam.com/backup/70/v ... eding.html recently so should I actually be having 2 backup respositories? like this...

Use Backup Job – to backup VMs to Primary Site Backup Repository (NTFS volume)
Use Backup Copy Job (using WAN acceleration) - to copy VMs to DR Site Backup Repository (NTFS Volume)
Use Replication Job – to Replicate VMs to DR Site ESXi/vSphere (VMFS Datastore) using DR Site Backup Repository as Seed

Seems like a lot of backups/copies of the same data, I'd have at least 2 copies of the VMs at DR, one in a VBK and one as the VMDK itself.

Thanks

v.eremin
Product Manager
Posts: 16134
Liked: 1315 times
Joined: Oct 26, 2012 3:28 pm
Full Name: Vladimir Eremin
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by v.eremin » Sep 24, 2013 11:35 am

1. Yep, Storage Migration results in CBT reset.

2. Replica seeindg is more about one-time operation that helps to reduce traffic sent trough the bandwidth during the initial replication cycle.

3. Such scenario guarantees required backup data redundancy. Though, it's up to you whether to follow these best practices or not.

Thanks.

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 24, 2013 11:44 am

1. Ah so if I moved a 2TB VM between datastores at the production site it would then have to replicate the whole 2TB (depending on compression and dedupe of course)? No way around this/workaround? Surely this must be causing people problems, especially on even slower WAN links.

2. Ok but in my design above would this actually mean better performance and mean I could replicate more data because I'd be able to use WAN acceleration rather than just replication and proxies at each end?

3. Is this more of a recommendation, ie. not essential? I suppose this means I could have more rollback points in the VBK and less snapshot rollback points on the DR VMs themselves?

Thanks for the help :mrgreen:

v.eremin
Product Manager
Posts: 16134
Liked: 1315 times
Joined: Oct 26, 2012 3:28 pm
Full Name: Vladimir Eremin
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by v.eremin » Sep 24, 2013 11:58 am

1. Nope, the reset CBT means that whole VM image will be read by VB&R; however, only changes will be transferred.

2. Not sure whether I completely understand you, but the backup copy jobs and replication ones are completely different and are used for different purposes. Also, the replica seeding procedure described above applies only to replication jobs and not to backup copy jobs. Likewise, WAN Accelerators are only utilized by backup copy jobs.

3. Yep, it’s recommended way to guarantee backup data redundancy. Though, as mentioned above, it’s up to you to decide whether to follow it or not.

Thanks.

CaptainFred
Enthusiast
Posts: 88
Liked: 2 times
Joined: Jul 31, 2013 12:05 pm
Full Name: Si
Contact:

Re: Backup and Replication to DR Site Design Proposal

Post by CaptainFred » Sep 26, 2013 10:57 am

Ok thanks.

Regarding the backup report email that is sent after the job is complete, am I right in saying that the “read” amount is the amount of changed data in the VM and the “transferred” amount is the amount it’s actually transferred to backup AFTER compression and dedupe?

I've just attempted a replication job of a VM from vSphere 5.0.0. I've selected my source and destination proxies but got the message "hotadd isn't supported for this type of disk". The destination proxy is installed on Windows Server 2012, could this be because technically hotadd isn't supported in vSphere 5.0.0 on Windows Server 2012? I can raise a support case if necessary about this but thought it might be simple issue if it's just the OS support. If I turn off the failback to NBD mode I get this error (couldn't find anything in the KB or forum for this and it doesn't have any snapshots):
Error: Client error: An existing connection was forcibly closed by the remote host Exception from server: Failed to open VDDK disk [[Datastore 1] SERVER1_R/SERVER1.vmdk] ( is read-only mode - [false] ) Failed to open VMDK. Logon attempt with parameters [VC/ESX: [192.168.1.1];Port: 902;Login: [****];VMX Spec: [moref=79];Snapshot mor: [79-snapshot-1];Transports: [hotadd];Read Only: [false]] failed because of the following errors: Failed to process [srcReplicateVddkDiskContent] command.

Post Reply

Who is online

Users browsing this forum: No registered users and 9 guests