Review the hardware requirements before installing Dynatrace Managed. Exact sizing of a Managed Cluster isn't always possible, especially in environments with growing traffic. While upfront analysis is useful, it's more important to be able to add capacity as your monitoring needs increase. Plan to scale along these dimensions:
Adding nodes increases how much your Managed Cluster can monitor. Keep in mind that a larger Managed Cluster also holds and processes more data. Cluster updates and other maintenance operations need more time to complete, and backups grow along with it. Plan for somewhat longer maintenance windows as your Managed Cluster grows, and scale out to gain monitoring capacity rather than to speed up existing operations.
Production environments require a minimum 3-node Managed Cluster for reliability and data redundancy. For details, see Multi-node installations.
The hardware requirements in the following tables are estimates based on typical environments and load patterns. Individual environments may vary. All values are per node. The storage columns show the disk space to provision: Transaction Storage covers 10 days of code visibility, and Elasticsearch covers 35 days of retention. Column definitions:
Max host units: The host units that one node serves. Multiply this figure by the number of nodes to get an initial estimate for your Managed Cluster.
Max Full-Stack OneAgents: The Full-Stack OneAgents that one node serves. Both columns are limits, so check which one your environment reaches first. Hosts with a lot of memory usually reach the host units limit first, while many small hosts usually reach the OneAgent limit first.
Minimum node specifications: CPU and RAM must be exclusively available for Dynatrace. Turn off CPU power-saving mode. CPUs must run at 2 GHz or faster. Hosts must have at least 32 GB RAM.
Transaction Storage: Dynatrace Managed distributes transaction storage across all nodes and doesn't store it redundantly. In multi-node Managed Clusters, Dynatrace Managed divides the total transaction storage by the number of nodes.
Long-term Metrics Store: In multi-node installations, Dynatrace Managed stores three copies of the metrics store. For four or more nodes, the per-node storage requirement decreases.
Keep the Long-term Metrics Store data below 2 TB per node. The disk sizes in the following tables include free space that Cassandra needs for compaction. Dynatrace Managed raises a warning cluster event when the data on a node exceeds 2 TB, and a severe cluster event when it exceeds 4 TB. Persistent storage growth degrades Managed Cluster performance, resilience, and availability, and causes issues when adding nodes. If you need more storage, add another node to reduce the per-node requirement.
Host units measure the RAM of your monitored hosts, not the amount of data those hosts produce. Two hosts of the same size can place different loads on your Managed Cluster, depending on how many processes, services, and requests run on them. Use the following tables for your initial sizing, then monitor your Managed Cluster and add nodes as your load grows.
A node can reach the recommended 2 TB of Long-term Metrics Store data before it reaches the host units listed here, so check both.
| Node size | Max host units1 | Max Full-Stack OneAgents | Peak user actions/min | Min node specifications | Disk IOPS | Transaction Storage | Long-term Metrics Store | Elasticsearch |
|---|---|---|---|---|---|---|---|---|
Micro | 50 | 25 | 1000 | 4 vCPUs, | 500 | 50 GB | 100 GB | 50 GB |
Small | 300 | 100 | 10000 | 8 vCPUs, | 3000 | 300 GB | 500 GB | 500 GB |
Medium | 600 | 250 | 25000 | 16 vCPUs, | 5000 | 600 GB | 1 TB | 1.5 TB |
Large | 1250 | 500 | 50000 | 32 vCPUs, | 7500 | 1 TB | 2 TB | 1.5 TB |
XLarge3 | 2500 | 800 | 100000 | 64 vCPUs, | 10000 | 2 TB | 4 TB | 3 TB |
The number of host units a monitored host consumes depends on the amount of RAM available on that host. A host with 16 GiB RAM consumes 1 host unit in Full-Stack mode, and a host with 64 GiB RAM consumes 4 host units. Because RAM is the only factor, a host unit doesn't tell you how much monitoring data a host produces. For full details, see Application and Infrastructure Monitoring. If you license Dynatrace through the Dynatrace Platform Subscription (DPS), Full-Stack Monitoring consumption is also based on host memory, measured in GiB-hours. To estimate host units, divide the total memory of all Full-Stack monitored hosts by 16 GiB. For example, 100 hosts with 32 GiB RAM each, monitored around the clock, consume about 28 million GiB-hours per year and correspond to about 200 host units.
See RAM requirements in Log Monitoring recommendations.
Dynatrace Managed runs resiliently on instances with 1 TB+ RAM/128 cores (2XLarge) and lets you monitor more entities, but this setup doesn't use the hardware optimally. Instead, consider using smaller instances (Large or XLarge).
In a Premium High Availability deployment, the node limit counts all nodes across both data centers: a 42-node Managed Cluster runs 21 nodes in each data center. Size both data centers symmetrically, with the same number of nodes and the same hardware, including disk storage, in each. For details, see Premium High Availability. The second data center holds a copy of the data. As a result, a node of a given size serves fewer host units than the same node size in a Managed Cluster with one data center. The following table accounts for this, which is why its figures are lower than in the Dynatrace Managed table. Multiply them by the total number of nodes, not by the nodes in one data center.
| Node size | Max host units | Max Full-Stack OneAgents | Peak user actions/min | Min node specifications | Disk IOPS | Transaction Storage | Long-term Metrics Store | Elasticsearch |
|---|---|---|---|---|---|---|---|---|
Large | 600 | 350 | 25000 | 32 vCPUs, | 7500 | 1 TB | 2 TB | 1.5 TB |
XLarge1 | 1250 | 700 | 50000 | 64 vCPUs, | 10000 | 2 TB | 4 TB | 3 TB |
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, consider using smaller instances (Large or XLarge).
To monitor 7,500 host units and up to 4,200 Full-Stack OneAgents with a peak load of 300,000 user actions per minute in a Premium High Availability deployment, you need 6 XLarge nodes. Place 3 nodes in one data center and 3 nodes in a second data center, each with 9 TB storage split across storage types.
Dynatrace Managed stores multiple types of monitoring data. Follow these recommendations:
Disk size can vary depending on usage. For example, in a 2-node Managed Cluster where each node has 10 TB and transaction storage on both nodes is only 1.5 TB, a third node needs a minimum of 1 TB: (1.5 TB + 1.5 TB) / 3 = 1 TB.
In a similar Managed Cluster where disks are 9 TB full, a third node needs a minimum of 6 TB: (9 TB + 9 TB) / 3 = 6 TB.
Anything below the calculated minimum is a misconfiguration. Multiple data sources, such as Session Replay, contribute to disk usage and can also trigger a misconfiguration.
Dynatrace Managed works with all common file systems. Use fast local storage suitable for database workloads. Dynatrace Managed also supports encrypted file systems when disk IOPS meet the requirements and encryption is transparent at the operating system level. Examples include Amazon EBS and Azure Storage encryption.
Requirements for Log Monitoring:
Additional recommendations:
Production environments require a minimum 3-node Managed Cluster for reliability and data redundancy. All nodes in a multi-node Managed Cluster must: