Managed hardware requirements

This topic explains the hardware for installing Dynatrace Managed. For other Managed requirements see Managed system requirements and Managed hardware recommendations for cloud deployments.

It's not always possible to provision nodes that are sized exactly right, particularly if your environment is subject to ever-increasing traffic levels. While it's useful to do upfront analysis of required size, it's more important to have the ability to add more capacity to your Dynatrace Managed cluster should your monitoring needs increase in the future. To leverage the full benefits of the Dynatrace Managed architecture, be prepared to scale along the following dimensions:

  • Horizontally by adding more nodes. We support installations of up to 24 cluster nodes.
  • Vertically by provisioning more RAM/CPU per node.
  • In terms of data storage, by being able to resize the disk volumes as required (for guidelines regarding recommended disk setup see below).

For cloud deployments, use the recommended virtual machine equivalents for Managed hardware recommendations for cloud deployments

The hardware requirements included in the following table are estimates based on typical environments and load patterns. Requirements for individual environments may vary. Estimates for specific columns take into account the following:

  • Minimum node specifications
    CPU and RAM must be exclusively available for Dynatrace. Power saving mode for CPUs must be disabled. CPUs must run with a clock speed of at least 2GHz and the host should have at least 32GB of RAM assigned to it.

  • Transaction Storage
    Transaction data is distributed across all nodes and isn't stored redundantly. In multi-node clusters, transaction data storage is divided by the number of nodes.

  • Long-term Metrics Store
    For multi-node installations, three copies of the metrics store are saved. For four or more nodes, the storage requirement per node is reduced.

    You should treat the 4 TB requirement for the XLarge node as the maximum acceptable size. If you need more capacity, consider adding another node. Plan your long-term metrics store for data being a maximum of 50% of your available disk space. In these terms, 4 TB of space would handle 2 TB of your long-term metrics store data. While stores larger than 4 TB are possible, they can make database maintenance problematic.

Dynatrace Managed

Node Type Max hosts
monitored
(per node)
Peak user
actions/min
(per node)
Min node
specifications
Disk IOPS
(per node)
Transaction Storage
(10 days code visibility)
Long-term
Metrics Store

(per node)
Elasticsearch
(per node)
(35 days retention)
Micro 50 1000 4 vCPUs,
32GB RAM
30 50GB 100GB 50GB
Small 300 10000 8 vCPUs,
64GB RAM
100 300GB 500GB 500GB
Medium 600 25000 16 vCPUs,
128GB RAM
300 600GB 1TB 1.5TB
Large 1250 50000 32 vCPUs,
256GB RAM
750 1TB 2TB 1.5TB
XLarge 1 2500 100000 64 vCPUs,
512GB RAM
1500 2TB 4TB 3TB

1 While Dynatrace Managed runs resiliently on instances with 1 TB+ RAM/128 cores (2XLarge) and allows you to monitor more entities, it's not the optimal way of utilizing the hardware. Instead, we recommend that you use smaller instances (Large or XLarge).

Examples

  • To monitor 10k hosts with a peak load of 300k user actions per minute, you need 3 extra large (XLarge) nodes with a storage of 9TB each split respectively to storage types.

  • To monitor 500 hosts with a peak load of 30k user actions per minute you need 3 small nodes with 1.3TB storage each split respectively to storage types. Alternatively, you can also use 1 medium node with a storage of 2.1TB.
    We recommend a failover set up of minimum 3 nodes instead of single nodes that are less resilient.

Dynatrace Managed Premium High Availability

Node Type Max hosts
monitored
(per node)
Peak user
actions/min
(per node)
Min node
specifications
Disk IOPS
(per node)
Transaction Storage
(10 days code visibility)
Long-term
Metrics Store

(per node)
Elasticsearch
(per node)
(35 days retention)
Large 600 25000 32 vCPUs,
256GB RAM
750 1TB 2TB 1.5TB
XLarge 1 1250 50000 64 vCPUs,
512GB RAM
1500 2TB 4TB 3TB

1 While Dynatrace Managed runs resiliently on instances with 1 TB+ RAM/128 cores (2XLarge) and allows you to monitor more entities, it's not the optimal way of utilizing the hardware. Instead, we recommend that you use smaller instances (Large or XLarge).

Example

To monitor 7500 hosts with a peak load of 300k user actions per minute in the Premium High Availability deployment, you need 6 extra large (XLarge) nodes - 3 nodes in one datacenter and 3 nodes in second datacenter. Each node with a storage of 9TB split respectively to storage types.

Storage recommendations

Dynatrace Managed stores multiple types of monitoring data, depending on the use case.
We recommend:

  • Storing Dynatrace binaries and the data store on separate mount points to allow the data store to be resized independently.
  • Not keeping Dynatrace data storage on the root volume to avoid additional complexity when resizing the disk later, if required.
  • Mounting different types of data storage on separate disk volumes for maximum flexibility and performance.
  • Creating resizable disk partitions (for example, by leveraging Logical Volume Manager [LVM]).
OneAgent opt-out

OneAgent self-monitoring is enabled by default, an opt-out installation parameter is available:

--install-agent <on|off>

Multi-node installations

We recommend multi-node setups for failover and data redundancy. A sufficiently sized 3-node cluster is the recommended setup. For Dynatrace Managed installations with more than one node, all nodes must:

  • Have the same hardware configuration
  • Be synchronized with NTP
  • Be in the same time zone
  • Be able to communicate over a private network on multiple ports
  • The latency between nodes should be around 10 ms or less.

We recommend that system users created for Dynatrace Managed have the same UID:GID identifiers on all nodes.

Avoid split-brain sync problems

While two node clusters are technically possible, we don't recommend it. Our storage systems are consensus-based and require majority for data consistency. That's why two node cluster is vulnerable to split-brain problem and should be treated as a temporary state when migrating to 3 or more nodes. Running two nodes may create availability or data inconsistencies from two separate data sets (single node clusters) that overlap and are not communicating and synchronizing their data with each other.