Windows iSCSI for Backups.

[BoomAM]

Hi,
Im trying to come up with a cheap reliable way to backup a variety of servers.

We have a server with 8Tb of storage in it purely for backups only.
My best idea for backups on the cheap is to create multiple iSCSI targets on the backup server for each server that needs backing up, then using iSCSI initiators to map the drives to each server, then whatever version of windows backup they use, do the relevant backup to that iSCSI drive. With the SQL server needing an extra iSCSI drive for SQL backups as 2008 Windows Backup(non R2) doesnt wipe the transaction logs.

Then, and this is where i need some input, i need a way to archive the targets off to external disks without causing problems. I would assume that theres some sort of script i could write that would tell the targets to 'dismount', allowing me to copy to an external disk, then remount? Or would VSS get around this for me...?

Thoughts & input very much appriciated.

Thanks.

 

[BoomAM]

I don't see the point in iSCSI at all now...

As said, only Windows Backup R2 can goto remote servers without using scripts (hard to monitor progress).

 

[BoomAM]

No, you said it doesn't clear down SQL logs, but SQL will do that if you use SQL to do the backups.

wbadmin can go to remote unc shares.

Yeah, so to get a complete backup, Windows Backup to take main image of system to iSCSI 1, SQL Maintenance Backup to clear logs/take extra DB Backup to iSCSI 2.

Wbadmin only does it via scripting, not via the GUI, so hence, i cant monitor whats going on, speed, etc;

Tempted to give DPM a try tbh, we can get it for free anyway...

 

[BoomAM]

Off the top of my head, at the moment, just one:

I take it that DPM works something like BackupExec, as in theres a application on a host server, and agents on the other servers. And together with the settings in the host server, they pull the data as required?

I'll probably have more depending on your answer to the above and as i begin faffing with it.

 

[BoomAM]

Correct.

Interesting....but i'd imagine that unlike BackupExec, when you select the shadow agent to use it actually either A) works or B) uses it.

Is there any faffing to set it up? Or is it literally as simple as BackupExec? (Install host, push out agents, set schedule/options, done?).
Or is there a lot of faffing a-la SCCM style? Where it takes a good few hours to get to the point where its even realistically usable. (i know SCCM isnt a backup product. ).

What about storage and secondary backups?
Can it be set to backup to a data drive on the host server, then at intervals, archive that data off to a USB disk or somewhere else?

Virtualisation support, HyperV only or can it interface with vSphere?
Or would the way around that just been to stick an agent in each of the VMs instead...?

Thanks.

 

[BoomAM]

Thanks,
I'll have to have a google for a good getting started guide and then give it a crack when the rest of the hardware turns up. Here's hoping it goes hunky dory!

 

[BoomAM]

Right.
In, configured (i think) and setup.
Few questions:

Am i correct in thinking that:
- Retention Range is when it decides to start overwriting data?
- Sync Frequency is how often it looks for changes to the server being backed up?
- File Recovery Point is when it backs up the file system?
- App recovery point is when it backs up the apps at an 'app level', like Exch & SQL?


Also:
When it backs up/does its app recovery point, does it clear Exch/SQL Transaction logs or do i still need to do them with a SQL Maintenance plan?
Because if it doesnt, then i may just change the frequency of those points from every hour to every day.

Is there a free way to get a USB disk to appear as a tape library to DPM, as i know its not supported by the software itself, so i assume some virtual tape library software would be needed.

Thanks in advance all.

 

[BoomAM]

Thanks for the quick reply!

So as an example, i have our Exch & SQL servers set to do a file recovery at 7 PM & 5:30pm reflectivity, with an app recovery being done at 9pm & 9:30pm respectively. And a sync time of every hour.
That should give me relatively good backups of both should it not?

With the sync's being every hour, the data required to be backed up for the file & app backups should be smaller shouldn't it?

I'll give firestreamer a google now.

Also, any ideas why the data drive for one of our SQL servers wont create a recovery point but everything else works fine? The same with a lot of our Exch mailbox databases...?

 

[BoomAM]

Odd.
Just had the server update itself, restarted it, and now DPM wont load as the service keeps crashing, and the SQL part of it wont start because of a login error! wtf!

##EDIT##
Fixed. Another remnant of my predecessor, GP refreshed and set the local login as a service setting to specific users. Once that had the DPM accounts added, all hunky dory!

 

[BoomAM]

The more often your syncs the smaller the amount of data transfered with each sync, but the file backup and the express full backup will still be the same size

Right.
So the sync will be smaller the more often it syncs, and allows a more granular restore between the other backup times if needed.
File level is a full backup at file level, always equal to the size of the data on the server.
App level is a full backup of the app, like Exch or SQL, but not the rest of the system.

Correct?

Fixed most of my errors. Just fixing a small verification issue on one of the file servers at the moment and jobs a good'en!

Quite impressed thus far.

 

[BoomAM]

Some more queries/questions.

Q1:
The recycle bin is reporting its using a ton of custom ammounts for each of the NTFS volumes that DPM has set up, to save space, is it ok to set them all to 1Mb? Or has DPM set them to what they are for a reason?

Q2:
I thought DPM, when the protection groups are set to 'auto grow', should 'auto grow' volumes to fit backups into.
Its started erroring about it all for some reason, even though there's another 3.8Tb free on the drive...

Q3:
Sync's are meant to be smaller the more often they happen...so why with 20Gb's of Databases is it transfering 32Gb each time it syncs...?

Q4:
On-the-wire compression: Does it make much of a difference?
Im just replanning the backup times so no two jobs run at the same time, if compression can make a big difference, it'd be handy...

 

[BoomAM]

Its starting to annoy me now.
Ive just uninstalled it and restored it to fix the reporting fault, and now it wont see the agents and the rollups wont install!

Just gonna completely wipe it off and start from scratch.

 

[BoomAM]

Hmm.
Mostly working again now.
Still got the reports problem.
But the disk space in use is less!? and for some unknown reason i cant get it to see a server at another site, it keeps saying its part of a cluster when its not?

 

[BoomAM]

Right, all in and back working, touch wood.
Just a few errors/issues:

1) Still errors with anything to do with reporting. Some user impersonation issue.
Fixed - Local security policy that was domain enforced. Just added the two DPM user accounts to the 'backup operator' group, and the 'backup operator' group to the GPO, job done.

2) Whats the best way to backup DPM itself? Am i literally stuck with that 3rd party software into fooling it into thinking a USB disk is a tape library, or is there another way around it?

3) Still cant get it to install an agent on a remote computer located across a site-to-site link, it thinks the server, a HyperV server, is part of a cluster, when its not.

Any ideas?

Thanks.

 

[BoomAM]

...snip...

The iSCSI Targeting software and the other parts of storage server are just normal installers now...

3rd Party, Tape library or Secondary server i'm afraid

Boulderdash.
Will Windows Backup not backup DPM?

No, that's a strange one, we have an HV cluster at both our sites, and a couple of stand alone HV server, and they all back up fine.

Its odd, as the agent installs manually fine on it, but DPM wont play ball. No blocked ports on the firewall, its very odd.

 

[BoomAM]

I guess it may do... but you would need as much stroage as you are using for DPM again to be able to back it all up.

Shouldnt be an issue, we're currently using 2.5Tb on the DPM server, and i have two 3Tb USB3 disks...

Got it working on Cluster, Non Cluster, Full Server, Server Core, all with firewall on, never had an issue.

By firewall i meant the ISA's at either end of the link. Both are passing traffic.
Its very annoying, as the fact that the link is limited to a specific bandwidth would mean that the scheduling features and agent throttling in DPM could be very handy.

Have you run setdpmserver after doing a manual install?

Yes.
I have a script that I've used on all the other servers that flushes DNS, installs the agent with the set server variable, and pipes the result to a log file.

 

[BoomAM]

So after the install you go to the DPM agents tab and add a new agent? What does it do? Any errors in the jog logs?

Yes, but when its 'adding' the agent/server it processes for a bit, then fails, saying it cant see the server irrc.

 

[BoomAM]

As more info, it keeps coming up with error 346.
WMI is on, Firewall on the client is off, and the ISAs are set to pass all traffic between the two machines.

Google searches dont lead to a lot of information

 

[BoomAM]

Have you got another machine in the remote location that you can use to test? Even an xp box or something, just to rule out it being a problem with the server.

Best i can come up with is wbemtest, works from a server at the same site to the server im trying to backup, but not from the main site (here) to the server im trying to backup, rpc errors.
Which is odd, as said, ISA is set to pass all traffic...