Dynatrace Managed cluster failover mechanism

[SB_12]

HI , someone please help me understand following excerpt from Dynatrace document :

"If you plan to distribute nodes in separate data centers, you shouldn't deploy more than two nodes in each data center. The replication factor of three then ensures that each data center has all the metric and event data. Also, for seamless continuity you need at least three data centers, one of which can fail. "

"Plan to deploy a minimum of three nodes per cluster "

Q. What is the importance of having two nodes or three nodes in cluster?

Thanks

Swati

[Radoslaw_Szulgo · Accepted solution]

That is correct statement. Let me explain it to you.

For a high availability deployment across multiple data-centers we currently support only 1 model with a following requirements:

• 3 data-centers
• low latency between DCs (~10ms)

Considering then we have the replication factor of three then ensures that each data center has all the metric and event data:

• The lowest deployment then is then 3 nodes split to 1 node per each DC. Having that you can afford of 1 DC failure to keep the cluster operating. However, 1 node failure destroys the DC.
• The recommended deployment is 6 nodes split to 2 nodes per each DC. Having that you can afford of 1 DC failure (in total 2 nodes out of 6 nodes lost) to keep the cluster operating. Additionally, you can lose 1 node in a DC to keep that DC running.

Having 3 nodes at each DC makes no sense e.g. (3+3+3), as losing 3 nodes out of 9 nodes breaks the cluster.

We are working now on a Premium High Availability solution that address above issues and limitations.