SnapProtect questions

[peluso]

Hi all,

You may have seen the latest announcement for our new product, SnapProtect.  Please ask your questions - we have technical expertise ready to provide answers and point to useful information.  I will post a few questions here as well to break the ice and open up the conversation.  For more information about the product, start with our product pages. http://www.netapp.com/us/products/protection-software/snapprotect.html

Best,

Terri

[bwood]

More on this one...

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.  <<< This is controlled under the covers by SnapProtect / DFM… I don't think you can tweak this.

I believe this behavior changed with OnCommand 5.x... the destination volumes names should match the source volume names.  From what I can tell, any duplications in volume names would be handled with an appending _X  ("myvol_1", myvol_2", etc).

[l_mitchell]

Thanks for all your replies. The naming question came about from the 'fan in' of multiple snapvaults, yes it does append the duplications with an '_number', and isn't an issue really, I keep thinking from a NetApp level rather than a CommVault level, I shouldn't care what the library storage is called as long as its there, its just I can see it on my NetApp storage and the naming convention bugged me!!! I think this is the hurdle people need to get over to get the best out of SnapProtect.

You mentioned iSCSI RDMs. Could you just clarify, you can do it with either physical iSCSI RDMs presented from VMware or is it in-guest MS iSCSI Software Initiator that you mean, or either?

Exchange VMs, would you protect it simply at a VM / VMDK level and use offline mining when required, or present luns to the VM and protect it as a physical? I'm guessing this is going to come down to how often you need to recover individual items. if few and far between then keep it in a vmdk / offline mining when neccessary is the way to go I guess, as again I'm going to probably need another proxy to offload the snapmining to?

Where do you perform a manual file level recovery if you didn't index the backup of a VM?

I have some more questions too: Libraries used for indexing. Would you have different indexes for say VMware VMs, NAS, Exchange, SQL etc or can you just use one or two or three indexing libraries? Someone mentioned having different libraries even down to the individual storage policies, keeping potentially different retention settings away from each other? I like the idea of keeping them on NetApp CIFS shares rather than on a lun, Any pro's / con's to these different ways?

The automatic updating features of the product, things like the cookbook say don't enable it, however the NetApp branded books online documentation for SP7 seems to suggest you can, Could you tell me whether it is?

The creation of two seperate backupsets for the same data, I think this is simply because the extended retention rules only really begin at weekly, I wonder if you can suggest to CommVault to include daily and hourly retentions and it would be as simple as that to make SnapProtect more NetApp friendly Also making Books online a bit more searchable for key items such as how to index a non indexed vm, or if I see a radio buttion and I don't know what it does i just want to search books online for it. I can't seem to do that currently. And CommVault have some sort of Stone Masons secret cult thing going on where they keep the field guides locked away in their technical reference library, I think you need to OEM a NetApp branded / cutdown Netapp centric version for partners that sell SnapProtect as I'd love to know whats in there? Theres a Commvault NetApp fieldguide from last year in the wild but is that the latest version? No idea, however I would love to know, purely as supplemental material to what I have already, or do you think your docs are better than those from Commvault now?

[bwood]

You mentioned iSCSI RDMs. Could you just clarify, you can do it with either physical iSCSI RDMs presented from VMware or is it in-guest MS iSCSI Software Initiator that you mean, or either?   <<< Either

Exchange VMs, would you protect it simply at a VM / VMDK level and use offline mining when required, or present luns to the VM and protect it as a physical? I'm guessing this is going to come down to how often you need to recover individual items. if few and far between then keep it in a vmdk / offline mining when neccessary is the way to go I guess, as again I'm going to probably need another proxy to offload the snapmining to?   <<<  Mining doesn't really change.  What changes is the ability to restore the database.  With VMDK approach you aren't using the Exchange agent so there's no seemless recovery of the database.  You could restore as flat files and then do something with that, but there's no automation.  With luns you can use the Exchange database agent which would allow you to restore the database.

Where do you perform a manual file level recovery if you didn't index the backup of a VM?  <<< The process is the same whether you indexed the VM or not.  When you browse the backup and select the option to do an individual file restore it will either list the files from the indexed backup or, if it wasn't indexed, it will build a temp index of that backup and then list the files.  While it builds the temp index you basically see "loading" on the screen until it's done and shows you the files.

I have some more questions too: Libraries used for indexing. Would you have different indexes for say VMware VMs, NAS, Exchange, SQL etc or can you just use one or two or three indexing libraries? Someone mentioned having different libraries even down to the individual storage policies, keeping potentially different retention settings away from each other? I like the idea of keeping them on NetApp CIFS shares rather than on a lun, Any pro's / con's to these different ways?   <<< I setup the libraries as NetApp CIFS shares and you can break it up however you like.   I probably wouldn't have a single media agent doing all of the work though. 

The automatic updating features of the product, things like the cookbook say don't enable it, however the NetApp branded books online documentation for SP7 seems to suggest you can, Could you tell me whether it is?   <<< Automatic updates for the CommServer needs to be disabled.  Updates for SnapProtect need to be manually pulled down from support.netapp.com as service packs are released.   However it should be ok to enable automatic updates on the clients as they should always pull updates from the CommServe's software cache.

The creation of two seperate backupsets for the same data, I think this is simply because the extended retention rules only really begin at weekly, I wonder if you can suggest to CommVault to include daily and hourly retentions and it would be as simple as that to make SnapProtect more NetApp friendly.  <<< Already done and being prioritized.

Theres a Commvault NetApp fieldguide from last year in the wild but is that the latest version?   <<< They are working on updating it.  It's from last year.

[BENPETZOLD]

I have a question to classic NDMP Restores with snapprotect: I have seen that commvault is providing a tool which is called ndmp restore enabler which provides the function to restore data from a ndmp tape to a windows host. My question is: Is this tool also working with the netapp snapprotect solution?

[bwood]

I have a question to classic NDMP Restores with snapprotect: I have seen that commvault is providing a tool which is called ndmp restore enabler which provides the function to restore data from a ndmp tape to a windows host. My question is: Is this tool also working with the netapp snapprotect solution?

Yes the NDMP Restore Enabler is included.

[BENPETZOLD]

Thanks for the quick answer. This are good news. If I remember right commvault mentioned that after a file restore of windows files the original ACL will be removed from the files. Do you know if this is still the case with the NDMP restore enabler in netapp snapprotect?

Thanks

Ben

[bwood]

Ben,

Per the docs...

Can my Restored Data lose its original characteristics?

Yes, it could. When restoring NAS data to a file system, you should consider how the data is restored. When data is restored to an environment from which it did not originate, the data may not always assume its original characteristics as the following describes:

Windows

• File and directory names with case differences will be treated like files and directories of the same name. This may cause one file/directory to overwrite another during the restore.
• Files and directories with names containing Unix-specific characters may have those characters replaced with a "_" character or have some other form of modification in the restored file/directory name.
• Windows allows 1024 characters for filenames, including the path. A filename, including the path, with more than 1024 characters will not be restored to the Windows computer.
• Files and directories that contain Access Control Lists (ACLs) or permissions are not restored.
• Links within files may not be restored.

Unix

• Files and directories that contain Access Control Lists (ACLs) or permissions are not restored.
• Links within files may not be restored.

[BENPETZOLD]

Thanks again, for the information! One last thing: Could you please tell me where I can find this document? We are planning at the moment how we can build our next storage and backup solution.

[bwood]

Try this link... I pulled the above from the FAQ tab on that page.

http://support.netapp.com/NOW/knowledge/docs/snapprotect/relsnap_protect90sp7/215-07144_A0_books_online_90sp7/books_online_1/english_us/products/nas_n...

[BENPETZOLD]

Thanks again for your quick help!

[l_mitchell]

Exchange VMs, would you protect it simply at a VM / VMDK level and use offline mining when required, or present luns to the VM and protect it as a physical? I'm guessing this is going to come down to how often you need to recover individual items. if few and far between then keep it in a vmdk / offline mining when neccessary is the way to go I guess, as again I'm going to probably need another proxy to offload the snapmining to?   <<<  Mining doesn't really change.  What changes is the ability to restore the database.  With VMDK approach you aren't using the Exchange agent so there's no seemless recovery of the database.  You could restore as flat files and then do something with that, but there's no automation.  With luns you can use the Exchange database agent which would allow you to restore the database.

I still haven't got round to testing recovery steps from within a VM / db in a vmdk yet which is my preferred route (chances of requiring this are minimal, but I'd like to know what my options are). You say restore flat files and do something with that....as in a SnapProtect process, or do you mean by recovery storage group in Exchange? What I'm' thinking is would SMBR be a vaild alternative if you are licensed for it / can you point it to the mounted flat files (since its basically Powercontrols rebadged), or is there a way for Exchange single instance recovery within SnapProtect for databases on vmdk which by the sounds of it there isn't, I initially thought that was what the offline mining thing did somehow, again maybe I'm comparing SMBR to this utility and getting confused...again. Thanks for all your help.

[bwood]

Offline Mining is similar to SMBR, but SMBR is a bit more powerful from what I understand.   Regarding recovery, since the db is in a vmdk you cannot use the Exchange agent to protect the db (because the agent cannot understand how to get from vmdk to NetApp volume snapshot - only the Virtual Server Agent can do that).  So in a vmdk you could restore the db directory to some location and use Offline Mining or SFSR against that restored data for granular recovery.  For db recovery you'd have to manually pull in the restored db files since the Virtual Server Agent doesn't understand Exchange and therefore cannot do an automated recovery of an Exchange db.

[frank_iro]

I understand SnapProtect uses only VSM and cannot use QSM. Is this true? Does that therefore mean that if I have a FAS2020 running OnTap 7.3.7 meant to be a replication destination for a FAS3210 running OnTap 8.0.3, it will not work? Archiving uses SnapVault, which is based on qtrees. Am I right to assume that the same scenario would work if the FAS2020 was used as an archive destination?

Suppose I flip it and have the FAS3210 as the replication destination, VSM would work, because you can SnapMirror from older to newer OnTap. But how would I then do a restore if necessary?

Does LREP work with SnapProtect? If I need to do an offline baseline transfer, how would I go about it?

Thanks.

[mwalters]

Hi Frank,

Yes, VSM or SnapVault only, so you are correct: for VSM the secondary ONTAP needs to be the same major release as the primary.   Yes, you could go from 7.3.7 to 8.0.3. 

A restore of *data* is fine: you sound like you are talking about reversing the SnapMirror relationship ?   That would not work, but then AFAIK SnapProtect does not offer a "DR & failback" feature: it is more about backup than DR. 

& yes, SnapVault cares little about ONTAP versions, so 8.blah to 7.whatever is fine 🙂

LREP does not intrinsically work with SnapProtect.....but I believe you could establish the relationship using LREP & then import the relationship into SnapProtect.

[frank_iro]

Thanks Walters,

You've mentioned that a restore of *data* is fine. How would I go about this? Is there a way to restore data without using VSM/SnapVault? I'm not too familiar with the SnapProtect interface, so perhaps my expectations are a bit different from the reality. Is there not a way to just select a backup and click "restore" and have it replicate data back?

As for importing LREP created relationships into SnapProtect, do you have a link to a guide on how to do this?

Thanks.

[l_mitchell]

Here are the notes I have about the SnapProtect import tool that you may need to be aware of:

• Import  tool creates NetApp clients, NAS subclients, and storage policies for existing OnCommand relationships
• Imports non-application datasets (import tool is for NAS only)

Imports external relationships except:

• Volume SnapVault relationships
• QSM relationships
• Fan-out relationships
• OSSV relationships

• Registers Snapshot copies for primary array
• Creates XML file which can be edited before final import
• Import does not preserve any throttle settings
• Does not import schedules, jobs, or Snapshot copies of external relationships

  Location…
     <Install Dir>\Base\SnapProtectImport.exe

Going off this info I would guess you would manually setup the relationship / use lrep from the cli, repoint the relationship normally, get it updating fine from the cil, then import into Oncommand, then use the SnapProtect import tool, assuming your situation doesn't fall into one of the above limitations of the importer. I hope this is of some use for you. If you do get to play around with this it would be nice to know your findings.

One of the things I notice when protecting a number of volumes with snapvault is it creates a 1-1 relationship ie every protected volume gets a snapvault volume created . I wonder if I created a fan-in relationship (which looks like it should be supported from the cli) import that into OnCommnand, then import that into SnapProtect would that work.

Another thing SnapProtect does is create a snapvault relationship for the volume, and one for each qtree. However it looks as the volume is ignored after that, what I dont know is does snapvaulting of a volume simply skip NetApp qtrees.

As for the restoring of data, as I understand it during the restore you select the copy precedence from which you want to restore, ie local snap is the default, and say you have an aux copy (snapmirror or snapvault) at 2 then choosing 2 from the restore advanced options will use that location. Something I would like answered around this, say its a snapvault destination, does it index on the fly or somehow link back to an original index assuming one was done. If you didn't do a cifs index initially, again if you want to restore will it build one on the fly, like for VMware VMs?

[l_mitchell]

Some more general questions.

Once it is licensed it looks like it is the full product, are you able to use commvault libraries as proper backup libraries (on NetApp CIFS shares) to be able to use the agents such as Active Directory or even just basic file server iDA for basic system state backups? Protecting the VMs is one thing but in my experience of protecting VMs you still need a good system state just in case. This also goes for Exchange / SQL. If I wanted to use NetApp SnapProtect as a traditional CommVault install, assuming it all resides on NetApp storage is this possible / legal.

How does the CommVault dedupe on its traditional type backup libraries and its aux copying of them to a secondary site compare to the way SnapProtect does it ie NetApp dedupe on the primary volumes with local snaps that are snapmirrored / snapvaulted. What I mean is how well does snapmirror transfers compare and also NetApp dedupe compare?

Are there any registry options we should be aware of, as the more I learn the more I realise a lot of things are registry tweaks.

A troubleshooting guide would be nice, as when things like snapmirror / snapvault break down, where should you start?? Also around this re: snapvaulting. As there are two snapvault relationships created. ie the volume and the qtree. If the qtree one fails but the volume one suceeds (as nothings changed in the volume) as far as SnapProtect is concerned it succeeds. Any general advice around snapvaulting troubleshooting and where you should start would be nice. Cli up or snapprotect backwards?

A couple of general things also from my experiences that may help someone else:

Protecting VMs with Exchange in VMDKs:

• to me it wasn't very clear that you need the file server and vss agents inside the VMs for it to truncate the logs (probably for SQL too), it sort of reads agentless to me, until you get into the nitty gritty of it. So you have client computers inside your CommServe that you don't do anything with. This is what got me thinking about the system states etc and trying to back that up, however I couldn't seem to be able to do so.

Scheduling of Auxilary jobs:

• doing a snapmirror then snapvault relationship? If you place a schedule against the snapmirror it will only do that step, whereas if I placed the schedule against the snapvault (copy precedence 3) it performed a snapmirror followed by a snapvault. So it seems you just need to put the schedule against the last copy operation. (in this scenario at least)

VMware Backups

• if you want to do hourlys, just do NAS backups, as for me the time it takes to take vmware snapshots takes an age! In the rare circumstance you are going to have to restore a VM just accept 5 additional minutes onto the time taken to export and mount the flexclone manually, you can clone it from snapprotect at least still!
• The VMware backups seem to moan from SnapProtect that all my VMs didn't quiesce. Yet from VMware, I only saw this every now and again (for known very busy servers). I think this is because the OS page file was on an excluded datastore, I also excluded the vmswp volume, however my money is on the page file volume.

[l_mitchell]

What is the procedure for troubleshooting auxilary jobs? I am seeing from SnapProtect a successful aux job, however upon inspection of the job that ran from NMC it shows that only part of the snapvault portion of the job was successful (the volume level one, that has nothing in it), but the qtree snapvault fails (the main part I need to work, with all the data in it!).

It is worrying that from SnapProtect that this is reported as successful.

From NMC and watching the console, it says that the 'source qtree does not exist' yet it absolutely does, I even use the /vol/... output from the previous console output at the cli which shows my qtree.

From an NMC point of view it is saying its non conformant too, so for SnapProtect, where do you start to troubleshoot this type of issue?

Can I just delete snapvault relationships from the cli and SnapProtect will create a new one via NMC? Or will it leave stale entries.

What will happen if I deleted the snapvault aux job, then recreated it? Would it create new vault relationships?

These were all native SnapProtect auxilary jobs that created the snapvaulting relationships.

The snapmirroring seems very stable, but seems to have fallen over at the snapvault side and I can't find out the procedure to troubleshoot it. It has been logged with NGS to no avail for two weeks now. So if anyone has any experience on this, please share with the community!

Thanks.

[bwood]

You can check the /etc/log/snapmirror log file on the destination array.  There's also a conformance.log file on the OnCommand (DFM) server.  It will be best to escalate this within NGS.

[jcbettinelli]

Hello,

Here are some additional questions:

• Is there a way to manage the naming that is generated in OnCommand/DFM for the Datasets and the Storage Services. I've just noticed that the Storage Services description include the name of the Storage Policy at the end, but nothing else. It would be nice if we could.
• What about the cleaning of DFM. When deleting a subclient for the NAS for example, it doesn't delete the dataset, neither the auxiliary copy. Is there a way to have this done automatically?
• Does the VSA agent support the VMware vCenter appliance?

Last question, which may be stupid (I may have missed something), but what is the purpose of having incremental or full backup, based on snapshots? Snapshots are not behaving like this.

Thanks.

Message was edited by: Jean-Christophe Bettinelli - Added vCenter question

[bwood]

#1  There is currently no way to customize the naming.

#2  Correct.  When you delete a subclient it will continue to manage the days retention for the storage policy copies.  The DFM dataset is not deleted.  The relationships are not deleted.  In order for it to remove the dataset and ultimately the relationships you would have to delete the entire storage policy.

#3  Good question.  Please email xdl-snapprotect@netapp.com on this one so product mgt can respond.

#4  Full vs Incrementals...  two answers to this.   (1) if you are backing up data to tape and you need incrementals to tape then you would need an incremental backup schedule established.  There would be no difference between a full and an incremental as far as the actual NetApp snapshot goes.   (2) When backing up NAS data incremental backups will perform the indexing phase faster than when doing a full backup.  So in a high file count environment incremental backups can help improve backup times by reducing indexing phase times.

Thanks!

Chris