HA Peers in different physical locations

Frequently Asked Questions
Post Reply
Tomas
Posts: 34
Joined: 15 Sep 2008, 15:57
Location: Clavister HQ - Örnsköldsvik

HA Peers in different physical locations

Post by Tomas » 25 Feb 2013, 10:33

This FAQ applies to:
  • Clavister CorePlus 9.x and cOS Core 10.x
Question:

Can we create a second Data center and “split” the HA-cluster between both physical locations?

Answer:

We do have some customers splitting their HA clusters. Having one node on each side of a city is usually no problem, we do however recommend having the cluster nodes in as close proximity to each other as possible as there is a big requirement regarding latency on the sync interface. If latency is higher than 20 milliseconds the sending cluster node will initiate a retransmit of the synchronization packet.

The heartbeats and sync traffic must not be “routed” (TTL=255 on departure as well as on arrival).

A common solution is to run a dedicated fiber cable for the sync interface. You can also solve it in other ways, but the connection they setup must be similar to a “direct cable” between the peers.

The bandwidth is highly dependent on the number of connections active and which other features are used, that are synchronized, as not all features are synchronized. Refer to the User Guide for details. There will be bandwidth peaks during reboot of a unit or similar.

Usually the packets are many and small, putting some strain on the NICs and intermediate equipment. The User Guide has details about the type of packets and their specific properties.

In short, a dedicated 1 Gbps link is usually what is needed, but it will vary with the traffic patterns.

Some care should also be put into designing the connection from each interface of the peers to its "common switch" so it mimics the default way of setting up an HA cluster.

Summary:

It is possible to place the clusters in different physical locations but it is not recommended as high latency synchronization is not something that is actively tested in QA, so there may be unknown behaviours.

It is also recommended to read through the following article about cluster setting adjustments: viewtopic.php?f=18&t=5744

Post Reply