Skip to main content

Configure custom t-shirt sizes for Lumi Enterprise

AI summary
Explains how to configure custom t-shirt sizes for Imply Lumi virtual compute pools using Terraform. Covers creating, modifying, and removing custom sizes on AWS deployments.

About AI summaries.
Early access

Virtual tier is an early access feature. If you don't have access to the virtual tier, contact your Imply support representative.

By default, Imply Lumi provides a set of predefined t-shirt sizes for virtual compute pools. If need different sizing than the defaults, you can define custom t-shirt sizes in Terraform. When you create a custom t-shirt size, you have the option to specify the instance type, node count, and query concurrency that best match your workload requirements and performance needs.

Custom t-shirt sizes are only supported for Lumi Enterprise deployments on AWS.

Prerequisites​

To configure custom t-shirt sizes, you need a Lumi user with the Admin role. For information on roles and permissions, see Manage roles.

Predefined t-shirt sizes​

The following table shows the predefined t-shirt sizes available in Lumi Enterprise:

SizeNameInstance type1Max concurrencyNode count
xsX-smalli4i.xlarge42
sSmalli4i.xlarge44
mMediumi4i.xlarge48
lLargei4i.2xlarge88
xlX-largei4i.4xlarge128
2xl2X-largei4i.4xlarge1216
3xl3X-largei4i.4xlarge1232

1 Instance types are AWS EC2 instance types.

Custom t-shirt sizes​

If the predefined sizes don't meet your requirements, you can create custom t-shirt sizes in Terraform. Custom t-shirt sizes replace the predefined sizes entirely; they are not appended to the existing list.

Create a custom size​

To create a custom t-shirt size, set the virtual_compute_tshirt_sizes property in main.tf:

variable "virtual_compute_tshirt_sizes" {
type = list(object({
size = string
instance_type = string
node_count = number
name = optional(string)
max_concurrency = optional(number)
}))

Each custom t-shirt size must meet the following constraints:

  • size must be a unique string value.
  • node_count must be an integer greater than or equal to 1.
  • max_concurrency must be an integer greater than or equal to 1. If you don't specify a value, it defaults to 4.
  • max_concurrency must not exceed the merge buffer limit for the selected instance type. Exceeding this limit prevents the virtual compute pool from starting.
  • instance_type must be one of the supported instance types listed below.

The following instance types are supported for custom t-shirt sizes, grouped by family. Each value shows the merge buffer limit for that instance size.

  • i3en family: xlarge (4), 2xlarge (8), 3xlarge (12), 6xlarge (24)
  • i4i family: xlarge (4), 2xlarge (8), 4xlarge (16), 8xlarge (32)
  • is4gen family: xlarge (4), 2xlarge (8), 4xlarge (16), 8xlarge (32)

The following example shows three custom t-shirt sizes:

virtual_compute_tshirt_sizes = [
{
size = "sc"
name = "small_custom"
instance_type = "i4i.xlarge"
node_count = 2
max_concurrency = 4
},
{
size = "mc"
name = "medium_custom"
instance_type = "i4i.xlarge"
node_count = 4
max_concurrency = 4
},
{
size = "lc"
name = "large_custom"
instance_type = "i4i.xlarge"
node_count = 8
max_concurrency = 4
},
]
}

After configuring a custom t-shirt size in Terraform, run terraform apply to apply the changes. The changes propagate to the Lumi UI automatically. New sizes appear grayed out in the Lumi UI while Lumi provisions the necessary resources.

Modify or remove a custom size​

To modify or remove a custom t-shirt size, update the virtual_compute_tshirt_sizes property in main.tf, then run terraform apply.

Keep the following in mind when modifying or removing a custom t-shirt size that is currently assigned to a virtual compute pool:

  • Removing a size that is in use causes the pool to stop serving queries and terminates all in-flight queries.
  • Modifying a size may affect the pool's ability to serve queries. For example, changing the instance type may cause the pool to stop serving queries.

To restore service, select an available size for the pool in Data management > Virtual compute pools.

Learn more​

For more information, see the following topics: