ONTAP Discussions

Snapview and flex clone questions

OLIVERLEACH
3,434 Views

Hi! New to NetApp and I have a couple of questions as I am looking at the product from a design perspective.

I have a scenario. I would like to replicate a vol which is on filer A to filer B, but I want to make sure the of the data integrity. Say the data on this volume was some MS SQL databases and I chose not to use database replication, but wanted to use replication at the array level.

Couple of questions if I may?

  1. First of all, would I need 2 Data ONTAP heads per filer, one in site A and one in site B or would I have a cluster and use the one OnCommand system manager? I need to plan for a loss of site A.
  2. Can I avoid clustering the NetApp filers?
  3. What happens if the latency is poor between the 2 sites maybe more than 15ms. I would typically want to replicate this using synchronous replication, not asynchronous to ensure data integrity. Could I use asynchronous and rely on SnapMirror to handle the integrity?
  4. Can I continuously snapshot using SnapMirror or would there be a schedule and if so, what is the minimum time I can schedule the snapshots at?
  5. If I use FlexClone, can I commit those changes back to the original volume created by SnapMirror?
  6. SnapVault seems to be a DR tool. This doesn't seem the right tool to use in this scenario. Is that correct?

Many thanks for helping if you can as I am sure this question comes up time and time again!

Cheers,

Oli-

1 ACCEPTED SOLUTION

aborzenkov
3,434 Views

1. It is possible to use one HA pair (MetroCluster) or independent filers with replication (SnapMirror). Exact choice depends on multiple factors.

2. Yes, but you lose most fault tolerance features.

3. Asynchronous replication in NetApp works with snapshots and it is guaranteed that you see either old or new state atomically. If all data is located on a single volume, snapshot is crash-consistent by definition. For multiple volumes you probably need application assistance to get consistent snapshots or use consistency groups.

4. SnapMirror works both synchronously and asynchronously. SnapMirror Sync ships data continuously, SnapMirror Async - based on schedule. Minimum time is 1 minute although it is not really recommended.

5. No.

6. SnapVault is backup tool. DR tool is SnapMirror.

View solution in original post

4 REPLIES 4

aborzenkov
3,435 Views

1. It is possible to use one HA pair (MetroCluster) or independent filers with replication (SnapMirror). Exact choice depends on multiple factors.

2. Yes, but you lose most fault tolerance features.

3. Asynchronous replication in NetApp works with snapshots and it is guaranteed that you see either old or new state atomically. If all data is located on a single volume, snapshot is crash-consistent by definition. For multiple volumes you probably need application assistance to get consistent snapshots or use consistency groups.

4. SnapMirror works both synchronously and asynchronously. SnapMirror Sync ships data continuously, SnapMirror Async - based on schedule. Minimum time is 1 minute although it is not really recommended.

5. No.

6. SnapVault is backup tool. DR tool is SnapMirror.

OLIVERLEACH
3,434 Views

Thanks for your response. Just one last question

Point 4

After doing some more reading, I see there is a product called SyncMirror. Does that give me real time synchronization, rather than snapshots? Or is it in fact just an extension of SnapMirror?

Point 6

Yes, I meant to say backup tool. Thanks for the clarification.

Thanks again

-Oliver

aborzenkov
3,434 Views

SyncMirror is mirroring between aggregates inside one head. MetroCluster is using it to mirror data between sites. Yes, it is real time synchronization. No, it is unrelated to SnapMirror.

OLIVERLEACH
3,434 Views

Great - many thanks. You have helped clear a few things up. I will say, after playing with NetApp and looking at the design considerations, it reminds me of SAN Symphony. It has similar functionality but WAFL seems to be the big difference and the way the

inode trees work. I especially like the way snapshots just copy the root inode and remap branches to any new blocks. That is great.

Thanks for your help on this,

Public