Any tips for initial replication?

  • 24 February 2016
  • 2 replies
  • 4072 views

Badge +5
We've got a 11mb pipe between sites that bursts to about 20. We have some PD's that we've set up with about 30 machines in them and turned on replication. Its not surprising that initial replication is going very slow and we have an estimated completion time measured in months.

Both sites are in production. I can't take on cluster down and bring it locally to the other site to get the intial replication done at high speed. Is there a white paper or procedure or even someone in here who's got some tricks to help us get replication moving? I was thinking of Starting a new PD, setting up replication, then adding one or two machines at a time until it's full and replicated. That seems pretty tedious and very time consuming.

Any help would be appreciated.

2 replies

Userlevel 1
Badge +5
Hello,

A couple of questions:
  • How much VM data you're attempting to transfer to the remote site?
  • Are there any bandwidth restrictions on the link?
  • Are you trying to replicate every snapshot remotely or only at intervals?

Replication is a low priority process and is designed to take place in the background, so the cluster is unlikely to utilize the full bandwidth of your link.

You can see on-the-fly bandwidth statistics by looking at the "Metrics" tab in the Data-Protection Table View.
Userlevel 7
Badge +30
Thanks for reaching out.

Looks like you're running 4.5.2, on Dell gear?

Can you open a support case with Dell and ask them to escalate it to Nutanix?

We can dive into your system and help you take a look at your specific setup, the associated metrics, and make sure the system is doing what it's supposed to be doing.


That all said, as John alluded to, this is all a function of getting data from A to B, so if you have a large amount of data doing an initial sync, over a relatively small pipe, its going to take a bit.

The one thing that would help maximize the situation is make sure that the remote site definition between the sites has network compression enabled, which will help stuff that pipe as full as we can get it.

Reply