Highlighted

Accelerator for CIFS

I've got a customer with HDS HNAS (Old BlueArc) and moving to Isilon for CIFS/NFS workloads.  

As an NBU customer (windows master and MS), NDMP was the protocol of choice for years, and has worked fine, albeit long Level 0's.  But with 7.7.x, now as we're digging into the accelerator, it seems like this is a great way to reduce the L0 issue.  

But we're wondering on the track log, how it determines the subsequent backups?  Does it walk the file system looking for the archive bit?  While it may move less data, I think the tree walk is going to be a time consuming process.  But the details on what happens on walking a CIFS/NFS mount is severely lacking in techincal details.

And what is the best practice to re-do a full backup?  60days?  90 days?  I think this is the reset track log option?  

 

Thanks!

 

 

 

9 Replies
Highlighted

Re: Accelerator for CIFS

Did you see that the plugin is available somewhere? I myself have all EMC devices and we are not able to take advantage of Accelerator. I know the Isilon is newer, but I am not seeing Accel_NDMP for it. Was wondering if you may have read differently elsewhere.

https://download.veritas.com/resources/content/live/SFDC/25000/000025228/en_US/nbu_77_hcl.html?__gda__=1466597733_ea7703d891aff3d22274701df36393d1

Highlighted

Re: Accelerator for CIFS

While there is an Accelerator plugin for NetApp, I thought you could use with it protecting CIFS shares?  That I simply had to map the share on a client, and go from there.  There were some caveats that NFS performed better or something like that..  I'm pretty sure I saw this as a use case/supported scenario?

 

Highlighted

Re: Accelerator for CIFS

Hey,

yes accelerator is only for netapp right now, and only supports disk based backups and that to dedupe disk i guess not sure.

 

But we're wondering on the track log, how it determines the subsequent backups? 

-->netbackup accelerator usally store track log on client, but in ndmp case it stores the track log on master and media server

And what is the best practice to re-do a full backup?  60days?  90 days? 

-->i recommend 60 days is a good pratice

--> and yes you can mount as CIFS or NFS mount and perform netbackup accelerator backups.

While there is an Accelerator plugin for NetApp, I thought you could use with it protecting CIFS shares?  That I simply had to map the share on a client, and go from there.  There were some caveats that NFS performed better or something like that..  I'm pretty sure I saw this as a use case/supported scenario?

and for all remaining questions

--> https://www.veritas.com/community/blogs/frequently-asked-questions-netbackup-accelerator-part-ii

Highlighted

Re: Accelerator for CIFS

Accelerator for CIFS share has been supported for a long time.

Herewith FAQ blog by former Veritas employee Abdul Rasheed: 

https://www.veritas.com/community/blogs/frequently-asked-questions-netbackup-accelerator 

Not sure if subsequent replies to CIFS-related questions will help.

The only other option is to get hold of a NetBackup PM as Abdul is unfortunately no longer with Veritas.

 

Highlighted

Re: Accelerator for CIFS

So on the incremental, does NBU still walk the FS?  

Highlighted

Re: Accelerator for CIFS

Acelerator avoids walking a FS because accelerator will query the FS meta-data instead, and build its own map of folders/files.  So, if accelertor is not enabled, or not possible, then yes the file system will still have to be walked.

Highlighted

I understand what the

I understand what the accelerator is to things that have an NBU client installed, and using the NTFS change journal and such.  The meta data exists from the FS itself, often using native OS hooks to a change journal that is already there....

But in this case, what is not covered in ANY technical detail is the mapping of CIFS/NFS mounts and how that track log is built out and maintained.  In your description, basically "magic," allows a closed OS (e.g. a NAS like HNAS or Isilon) to have metadata available to NBU to own the map of files/folders.  

What hooks specifically does NBU use to find changed files within a NAS FS?  I'm dealing with a NAS that has over 400TB of data with some shares having over 50 Million files.  NDMP is not fun.  At the same token, no one can give any technical detail how accelerator works on a CIFS/NFS.  

Re: I understand what the

Shame we never got a reply to this.  Teiva-boy asks exactyl the quesiton I'm thinking!

Is it jsut walking the FS again?  Or is there a hook?  Would love to know...

 

Highlighted

Re: I understand what the

Although not a direct answer and I am not sure if NBU does "NDMP Accelerator" this way, but here is a rough idea how it's been done in Avamar in the last 10-ish years for NDMP clients: a backup client (in Avamar it sits in a special box called NDMP accelerator, in NBU this should be a Media server) runs NDMP Tape Server for 3-way NDMP - because NDMP does not support backup levels beyond L9, the client operates with L0 and L1 only. When first backup is done, Avamar client generates all relevant cache files on the client (I'd assume at this phase NBU will do the same in the catalog) and for all subsequent backups requests L1 backups but does not transfer all files but rather looks for files from the list provided by the filer for L1 backup and checks their metadata - this does not require a walk through the file system. Once the backup is performed, Avamar marks this backup as L0 - because at the back end it will reassemble it to L0 and the cycle continues.

I'd expect NBU to do something like this.

Now here is a potential problem. For many filers I seen - the workload to generate the list for L1 is probably more resource intensive task than just copy all the files again. And I seen more frequently than otherwise, unfortunately, HNAS performing L0 faster than generating a list for L1 when millions of files are involved.