Backup or Replicate

[ChrisMoon]

Hi,

We've got two identical ESX4 hosts with local storage and approximately 10 VM's in production on one of the hosts.

We'd like to use the second host as a warm-backup environment so that if there are any problems with the first host, we can simply start-up the VM's on the second box.

I've assumed that the best way to do this is via replication, rather than a backup job?

Just to be clear, we've not worried about backing up the contents/config within the VM as we're using Backup Exec on each VM, this is to ensure that we have a relatively recent copy of the system should there be a hardware failure in the primary host. In the event of this failure, the secondary would become the primary host, and once the original host is corrected it would be setup as the secondary.

Thanks,
Chris

[Vitaliy S.]

Hello Chris,

It's up to you to decide your backup/replication strategy. However, if you want to have VMs that could be easily and quickly started when anything happens to your primary host - replication is the best choice, performing failover operation using Veeam Backup and Replication console will start your replicas immediately, in this case you'll have less downtime for your production environment.

Thank you.

[Gostev]

Chris, this comparison between backup and replication should help you making the best choice for your environment.

[dkvello]

Chris, this comparison between backup and replication should help you making the best choice for your environment.

Could we have a third option ? Read once/Write twice ? A combination of Backup and replica where the Data read (in VCB/SAN mode) would be sendt both to a replica destination and a Backup (local disk) destination in the same job ?

[Gostev]

This should be possible for us to add, but this will slow down backup to the speed of replication (which could be quite slow in case of WAN replication). Is that OK?

[dkvello]

This should be possible for us to add, but this will slow down backup to the speed of replication (which could be quite slow in case of WAN replication). Is that OK?

Well, as long as it's faster than doing two seperate jobs
In many cases the customer want's a replica to a secondary site (that is within fiber distance 2->8GB) for DR/DP reasons as well as a backup (VCB/SAN) to be able to do fast file level restore and to be able to move data to tape.

[Gostev]

Don't get me wrong, I totally love the idea - just wanted to make sure you understand the implications

[dkvello]

I'm asking this because I want to keep trafic off the Production storage. Doing a replica and a backup of the same VM's would mean I'd have to the same read job twice (more or less).
Would it be feasible to do a backup to disk (VCB/SAN) and then have a function to update the replicas from there ? The replicas are usually on a separate storage system (most FC).

[Gostev]

How I am seeing this is Veeam Backup pulling data locally, and then sending it to multiple targets (backup storage on file server, and replica on remote ESX). So the data will only be read once from your production storage. Is this what you were thinking?

[dkvello]

How I am seeing this is Veeam Backup pulling data locally, and then sending it to multiple targets (backup storage on file server, and replica on remote ESX). So the data will only be read once from your production storage. Is this what you were thinking?

Yes

I figure most customers will have their Backup on VCB/SAN and backup to a local Storage (DAS). They want this to be as fast as possible.
Once they've moved the data off the Production environment as fast as possible (and only once) They're no longer bound to strict Backup/Replication windows.

Then they want to do one or more of the following from the Local Backup/VCB/Veeam server:
a) Replicate the Backup-server Data to a remote location so that a second VCB/SAN/Veeam server can restore VM's to a secondary/hosted VMware environment
b) Backup the Veeam Server to Tape (with f.eks. TSM for long term storage) so they can do a complete restore of the whole Backupserver with data in case of an emergency or audit.
c) Update Replicas on a secondary Storage system, either Locally (FC/SAN) for Near Realtime Failover/DR or off-site for offsiete Failover/DR (Slow lines)

[ChrisMoon]

Chris, this comparison between backup and replication should help you making the best choice for your environment.

Thanks for this Gostev, it's confirmed what I suspected, replication is right for us.

[cag]

[Merged into existing discussion]

If you can run from a backup why would you choose to backup or replicate a VM.
Whats the difference between backup and replication ?

[Alexey D.]

Hello Conrad,

Please refer to the table in the post above, it answers your question.

[Gostev]

This is actually 2 year old discussion, but still mostly good. Most up-to-date explanation is available in the sticky FAQ topic. Thanks.

[tsightler]

[Merged into existing discussion]
If you can run from a backup why would you choose to backup or replicate a VM.
Whats the difference between backup and replication ?

Running from a backup using Instant Recovery is, as Anton might say, "using a spare tire", and really, one of those pitiful little donut spare tires that are rated for like 50 miles. It's a temporary solution to an emergency. Surebackup Instant Recovery is quite awesome, but don't expect to get much IOP performance from a PowerNFS backed VM. For applications that need seriously quick recovery, with reasonable IOP performance, and a permanent failover to a DR site, replication is the correct answer.

[larry]

We use four options for Veeam replication ( well 3 but 4 soon ) does not include the SAN real time replication.

Replication to spare SAN ( High cost no negs )

Replication to low cost SAN ( low cost very good performance for a vm or two ) just installed a 4 tb iscsi for under 2k

Replication to internal esx server raid. ( low cost, very good performance, can’t vmotion, backups flow though ESX server, no foot print ) Our new hp servers have room for 8 raided - 2.5 inch disks anyway.

using Instant Recovery ( low cost – no cost, don’t have the performance numbers will be testing soon ) will use this for some readonly data we want online during a DR

Our main reason for switch more to replication from backup is the need to have some vms up quickly during a SAN failure or DR. Also I need to have a zero day virus plan, I let Veeam replicate with no write rights to any SAN, the esx servers will do the writing. This protects my backups from any Windows virus. Be sure in have at least nx2.5 for space for any replicate even with SAN snap shots, for the worst case where a virus wrote 100 percent new bad data to the disk.

[tsightler]

Also I need to have a zero day virus plan, I let Veeam replicate with no write rights to any SAN, the esx servers will do the writing. This protects my backups from any Windows virus. Be sure in have at least nx2.5 for space for any replicate even with SAN snap shots, for the worst case where a virus wrote 100 percent new bad data to the disk.

Yep, this is a very good idea and it's one of the reasons we backup to Linux targets. For the most part we use low costs iSCSI SAN's, 20-32TB's in size, front-ended by a physical, hardened Linux box. Veeam writes the backups files using an unprivileged user account.

After the recent issue with B&R 5.0.2 where Veeam silently deleted our older restore points, we tightened the security of our backup files even further by writing a script which sets the "immutable" flag on the backup files until 1 day before their retention is set to expire. That way even Veeam can't delete the files accidentally.

[bc07]

Windows Virus in what scenario? That the Windows virus deletes the backup files from the backup destination? I don't know how much data loss you can afford in your company but we are fine with tape backup and would loose maybe 1 day of data in case of a virus deleting/damaging data on the servers and backups on disk. Only downside with backups is of course as you mentioned it takes time to restore instead of just turning on replicated VM's. Having more than twice the capacity of well performing storage to house live and replicated VM's is a bit expensive.

[larry]

We are now up to 6 tapes a day, one tape failure a week: user, media, software or hardware. We do one restore a week, takes hours. We no longer want to ship tapes, too much risk. Just running out of time for tape and besides tape sucks. 2tb of tape, $80 2tb hard disk - tape drive $3500, low cost NAS or iscsi $2000
The only thing tape gives me beside a head ache is off line storage.
The Windows virus I am planning for is one that effects ALL windows units and their data in the enterprise. I work at a bank in this is just normal planning.