Synchronous Dual Data Center (SDDC)

Sergo
Padawan

Hi everybody!

Exasol 7

We are trying to reproduce SDDC as it is described here: https://docs.exasol.com/planning/business_continuity/sddc_details.htm

How do we configure so that Master Segments reside in DC1 and their Slave Segments in DC2?

We cannot find any documentation on this topic.

1 ACCEPTED SOLUTION

Accepted Solutions

exa-Nico
Community Manager
Community Manager

Hi @Sergo - 

You are correct, this is not documented. AFAIK, it's not possible to configure on which nodes the slave segments reside in Exaoperation, rather, you have to perform certain actions in a specific order. I've opened an internal ticket to add this to our documentation. 

In the meantime, I would suggest to contact Exasol Support - they should be able to help you further. 

Sports Enthusiast. Database Guy. Member of Team Exasol.
Having trouble? Just let me know!

Please Give Kudos if you find this post useful and also mark this as Accepted Solution if this is the required solution

View solution in original post

6 REPLIES 6

exa-Chris
Community Manager
Community Manager
Good Morning, let me find someone to reply today. Rgds Christian
Connecting Customers, Partners, Prospects and Exasolians is my passion. Apart from that I cycle, listen to music, and try to understand what all those technical discussions really mean...

exa-Nico
Community Manager
Community Manager

Hi @Sergo - 

You are correct, this is not documented. AFAIK, it's not possible to configure on which nodes the slave segments reside in Exaoperation, rather, you have to perform certain actions in a specific order. I've opened an internal ticket to add this to our documentation. 

In the meantime, I would suggest to contact Exasol Support - they should be able to help you further. 

Sports Enthusiast. Database Guy. Member of Team Exasol.
Having trouble? Just let me know!

Please Give Kudos if you find this post useful and also mark this as Accepted Solution if this is the required solution

View solution in original post

mwellbro
Xpert

Hi @exa-Nico , could you post an update here ( or a new topic in the forum ) once the "add this to our documentation" is done ?
From the current documentation I´d have guessed that the slave segments always reside on the neighboring nodes but that would contradict what was shown with the two data centers in the SDDC article - color me intrigued 😄

Cheers,
Malte

exa-Nico
Community Manager
Community Manager

I'll try to remember! In general yes, the slave segment is on the neighboring nodes, but if you do these actions in a specific order when setting it up, they are mirrored on the other nodes. 

I don't know too many details on how it exactly works yet actually 🙂 

Sports Enthusiast. Database Guy. Member of Team Exasol.
Having trouble? Just let me know!

Please Give Kudos if you find this post useful and also mark this as Accepted Solution if this is the required solution

Sergo
Padawan

Here is the answer from Support:

Regarding private or public VLANS, both data centers will have the same private and public VLAN. The management node will be a virtual machine which is synchronized between both data centers. For example:

If DC1 has 3 nodes and DC2 has 3 nodes
DC1 will consist of 2+1, master data will be written on nodes in DC1
DC2 will also consist of 2+1 with a database that is not running, the nodes receive redundancy copies from DC1
Management node runs on VM host, management node will be started on VM host in DC2 incase DC1 goes down, and the database of DC2 will be started.
As for the configuration of storage, both databases will use the same storage volume. The storage has it's slave segments in the secondary DC. This means that you will configure it as redundancy 2, select all data nodes for both databases and select 2 for master nodes which means the other two data nodes will be the slave segments. Please take a look at the snapshot that is attached on the Exa-ticket.

Charlie
Xpert

Hi,

 

maybe I can help with this.

 

We have been using this kind of setup for 6 years now.

 

What you might consider in this kind of setup is a sort of naming for the nodes in your primary data center and your secondary data center.

 

We went for n11,n12,nXX for the primary data centers and n111, n112, n1XX for the secondary data center.

 

I will just use this in the following:

  • Primary data center: n11,n12,n13,n14 (spare node)
  • Secondary data center: n111,n112,n113

 

Now we create a volume:

Redundancy: 2

Node List:

  • Add n0011,n0012,n0013 to list
  • Now add n0111,n0112,n0113

Number of Master Nodes: 3

 

The volume will look something like this:

Charlie_0-1637136972083.png

Basically what happens is that we have 6 nodes in a list with 3 master nodes. So the first 3 nodes in the node list of the volume will be the master nodes and the next 3 nodes in the list will be the nodes hosting the slave segments.

 

The order of the nodes in the create volume node list does is very important.

 

But now you only have a volume that stretches over 2 data centers and nothing more.

If you really want to be able to start the database in the secondary data center that's another story. If you are interested I can post this as well.

 

Please keep in mind that for a high DML performance the latency between data centers should be <= 150 µs and you should monitor the bandwidth.