Hi all, hope you can help with my understanding here, not that I'm in the middle of deploying it or anything never seeing the product before!
If I want hourly snapshots kept for 48 hrs for example I have to create a seperate storage policy / backupset, OK I get that. It would be nice if I could do something in the extended retention rules for hourly and dailys and not just weeklys / monthlys / yearlys but its a Commvault product out of your control.
Anyway so then if I want two weeks worth of dailys kept I create another storage policy / backup set and associate the same volumes to the subclient for backup. Correct me if I'm wrong here! Still all on the same local volume.
OK so now ideally I want hourly snapmirrors then a daily snapvault. If I use my daily storage policy / backupset for an hourly snapmirror it doesn't work as from Commvault it sees the snaps as unique, and that there aren't other snapshots on the volume. So I have to do a daily snapshot followed by a daily snapmirror followed by a daily snapvault.
And if I use the hourly snap storage policy it'll create a seperate snapmirror, so thats out of the question.
Some additional info. I have a NCommServe backing up NetApp CIFS shares, and a proxy sever setup for VMware, using datastore affinities for the backup selection. All
OK so the stage is set....!
The main question am I doing this right, does this sound like the right way to go about things?
Can I use snapmirror.conf from the NetApp side to keep the snapmirror relationship thats understood by the SnapProtect daily storage policy to update the relationship hourly without breaking anything inside snapprotect? So I will then get my hourly snaps created on the primary across to DR. I'm not bothered about Snapprotect knowing about those on the DR snapmirror.
If I protect multipe volumes (not qtrees) with a snapmirror aux copy to DR followed by a snapvault of the mirror where is the naming convention controlled? DFM? Is it the datasets I created? Not tried playing around with this bit yet, however if someone has an answer rather than trial and error that would be my preferred option. Any best practices on what you do to control snapvault destination names. Basically I'll have multiple volumes going into a single snapvault destination (fan in? see I am paying attention!) In all the demos it looks to be the sim where you only have a single controller, what I have done is rather than saying secondary / tertiary. I have called them drfiler1_aggr0_50tb / drfiler2_aggr0_50tb, and have the snapmirrors for filer1 going to drfiler1 and filer2's snapmirrors going to drfiler2 in case there ever was a DR situation (probably never but hey) theres some load balancing of the resources of the controllers. I then have set the snapvaults to go to the alternate drfiler, which is probably slightly excessive, but then the snapvault is on a different aggr to its snapmirror for additional resiliency. I have plenty of space to play with.
Also I see the individual qtree protected and the volume protected as two seperate snapvault relationships. Is this normal, I'm not seeing data being copied twice per volume / unnecessary workload on the filer am I? Not played with Snapvault extensively enough to know what I should be seeing, as normally you protect qtrees, but I'm not specifying what qtree to backup inside snapprotect.......or can I / should I be doing so???
Whats the best way to schedule a snapshot / snapmirror / snapvault job. Ideally I'd like it to just do one after each other as the previous one finishes. If this is an option and I'm just missing on where it is please let me know. Or is it just a case of scheduling each piece of the job manually and hoping they dont overlap. Again schedule policies baffle me a bit, should I be using them, it seems I need schedules to create a schedule policy? Also can I only kick off a primary snap by scheduling each subclient? It seems that way, the documentation seems to do it this way but doesn't say that is the case?
If backing up VMware (with the Primary snap) the Enable Granular Recovery option....I thought by unchecking that it would avoid the excessive flexcloning and mounting of VMs inside VMware but oooh no! Is what I'm seeing inside VMware some sort of basic verify? What is the granular recovery actually doing, is it indexing the vmdk once its mounted and been tested? What is the overhead for this? Can I create an hourly job that doesnt do this (dont say configure NAS snapshots!). Also what is the other option of Create Backup Copy Immediately, with Enable Granular Recovery for backup copy.......is this to do with backup to tape or something?
SQL in a VM best practices. I'm struggling to understand how the SQL bit works, not being much of an SQL guy. Anyway would you present RDMs (either vmware pRDMs or via vm guest iscsi maybe?) and treat it as a physical host, or do you leave it inside the vmdk and some sort of other magic happens to backup the DB? Do I need another proxy for this? I'd be very interested to understand how people have gone about this. Also the SQL server I'm talking about is currently a physical cluster with netapp luns, and as far as I can see I'll need a proxy anyway right? Something I am not looking forward to anyway!
Exchange seems to be simple enough, my scenario I have exch2007 and two mail stores (no CCR or anything like that thank god). I am going to dump all exchange VMs into its own NFS volume and protect them all in a single policy. For things like the individual mail restore options work do I just install all the agents for exchange server / mining etc on each mail server and that should be ok? Any gotchas / real world best practices you can share?
I have been through all the NetApp university online courses / have the TR docs / POC cookbook and am trawling through the web / online documentation, but its very slow so far as its a pretty complex beast, so any comments and suggestions are most welcome. Just trying to find other peoples example storage policies for some sort of reference seem to be impossible, it would be nice to see some scenarios where a SnapManager product did it this way, this is how you'd do it in SnapProtect.
Thanks guys.
