Org Domains

Always state these, in any answer drawn from this file:

• Trusted access for observabilityadmin.amazonaws.com must be enabled in AWS Organizations first, and because that is an organization-wide change you MUST ask the customer before enabling it — even if they already said to proceed.
• create-domain-for-organization is management account only; a registered delegated administrator can do everything else but not that. An account-scoped Domain on the management account blocks an org Domain and must be deleted first, which is the customer’s decision.
• A Domain grant is ADMIN only and reaches every Space in the Domain, including future ones. Prefer IDC_GROUP; do not use IAM_ROOT unless asked.
• Teardown order: every Space first, then the Domain, from its home Region, using delete-domain-for-organization (delete-domain rejects an org Domain). Deleting the Domain removes its grants, so do not revoke them first.

Contents

• Which path
• What you get when setup completes
• Prerequisites
• Step 1 — Interview the customer first
• Step 2 — Enable trusted access
• Step 3 — Create the domain access role
• Step 4 — Create the Domain
• Step 5 — Create the Spaces
• Step 6 — Grant Domain access
• Verifying the setup
• Cleanup
• Troubleshooting
• Security considerations
• Additional resources

Which path

The two are mutually exclusive on the same account. A Domain of either scope already on the management account blocks creating the other.

Who may call what differs across the org operations. Establish the caller’s account up front and match it against this table before calling anything:

Constraints:

• You MUST confirm the caller is in the management account before attempting trusted access or create-domain-for-organization. A delegated administrator can do everything else but neither of those.
• You MUST NOT use create-domain or delete-domain on an org Domain. Those are the account-scoped operations and the service rejects the mismatch.

What you get when setup completes

• A domain ID shaped d- followed by up to 25 lowercase alphanumerics, a domain ARN, and a domain endpoint URL derived from the domain name.
• The Domain also reports the organization ID and the owner account ID, which is the management account.
• A Space per member account per Region.
• One or more domain access grants, each with a grant ID (a UUID), naming the principals who administer the Domain.

An org Domain ARN uses the resource type organization-domain, where an account-scoped Domain uses domain. That difference is how you tell them apart in a listing.

Quotas and scope rules to state up front:

• One Domain per organization, keyed on the management account.
• One Space per account per Region, unchanged from the account-scoped path.
• A domain access grant applies to the whole Domain. It conveys access to every Space in the Domain, present and future — it is not scoped to one Space or one account.
• Domain access grants carry exactly one permission and it is always ADMIN. There is no read-only or scoped variant at the Domain level; narrower access is granted per Space.

Constraints:

• You MUST tell the customer the domain name becomes part of the endpoint URL before they choose it.
• You MUST tell the customer that a Domain-level grant reaches every Space in the organization’s Domain and is ADMIN-only. A customer who expects it to be limited to one account or one Space has misunderstood the blast radius.

Prerequisites

• Trusted access for CloudWatch Observability Admin enabled in AWS Organizations. Step 2 checks and enables it. Without it every org operation fails authorization, and the failure does not name the missing prerequisite clearly.
• A caller in the management account for trusted access and Domain creation.
• An IAM role in the management account to pass as domainAccessRoleArn. See Step 3.
• For an Identity Center Domain, the instance ARN. The Domain must be created in the instance’s primary Region; the service verifies this.

Region rules:

• An org Domain supports Spaces in Regions other than the Domain’s own.
• If the Domain uses Identity Center, a Space’s Region must also be one the Identity Center instance is replicated to.
• Not every Region is available for Spaces under a given org Domain. Confirm rather than assume, and read the rejection message when one comes back.

Constraints:

• You MUST resolve trusted access before doing anything else. Every other failure mode in this document is easier to diagnose once it is ruled out.
• You MUST NOT promise a Space in a given Region under an Identity Center org Domain without confirming Identity Center is replicated there.

Operations you will call

Every operation runs through the aws___call_aws tool as an aws <service> <operation> CLI command, with the operation’s inputs passed as CLI flags (--kebab-key). Policy documents appear below as standalone JSON blocks — pass each as the corresponding flag value (--assume-role-policy-document).

aws organizations operations:

aws cloudwatchomni operations:

aws iam operations:

aws sso-admin operations:

aws sts operations:

Report the caller identity before starting, and confirm the account is the management account:

Step 1 — Interview the customer first

Ask these in one message and wait for answers. Do not ask piecemeal, and do not default silently.

1. Region for the Domain?
2. Domain name? It becomes part of the endpoint URL. Lowercase letters, digits, and single hyphens between them; 3–63 characters.
3. Authorization provider — IAM, Identity Center, or both?
4. If Identity Center: the instance ARN. Also ask which Regions Spaces will be needed in, so replication can be checked before promises are made.
5. Domain access role — should the agent create a new role in the management account, or will the customer supply an existing role ARN? Recommend creating one.
6. Which member accounts need Spaces, and in which Regions? Also ask whether those accounts will create their own Spaces, or whether the management account should create them on their behalf.
7. Who administers the Domain? The principals to grant ADMIN to in Step 6.

Confirm the choices back in one line, then execute.

If the customer does not know their Identity Center instance ARN, look it up for them rather than sending them away to find it:

Each entry carries the instance ARN and its identity store ID. The public reference for this operation is sso-admin list-instances (opens in a new tab).

Constraints:

• You SHOULD recommend creating the role rather than reusing one. Unlike the account-scoped path, the service actually attempts to assume this role at create time, so a role with the wrong trust policy fails the create outright.
• You MUST ask which Regions Spaces are needed in when the Domain uses Identity Center. Discovering a Region is unavailable after the Domain exists means the Domain has to be recreated elsewhere or the Region abandoned.
• You MUST tell the customer in this same message that a Domain-level grant reaches every Space in the Domain, so they choose the Step 6 principals accordingly.

Step 2 — Enable trusted access

Omni’s organization operations depend on CloudWatch Observability Admin having trusted access in AWS Organizations. The service principal is observabilityadmin.amazonaws.com.

Check first

The response carries EnabledServicePrincipals, a list of entries each with a ServicePrincipal and a DateEnabled. Look for observabilityadmin.amazonaws.com among them. This call works from the management account or a registered delegated administrator. The public reference is organizations list-aws-service-access-for-organization (opens in a new tab).

Enable it if absent

Ask the customer before enabling — this changes an organization-wide setting. Present the enable-aws-service-access command and get the customer’s explicit confirmation before running it; do not enable trusted access as an implied step of the overall setup, because it affects every account in the organization.

This call is management account only and returns no body. Re-run the check to confirm the principal now appears.

Constraints:

• You MUST ask before enabling, and you MUST ask even when the customer has already told you to proceed. It is an organization-wide change affecting every account in the organization, their administrator may need to authorize it, and “go ahead” on the overall task is not consent for this specific change. Surface the command and wait.
• You MUST use exactly observabilityadmin.amazonaws.com. Do NOT use telemetry-pipelines.observabilityadmin.amazonaws.com — that is a different principal for a different purpose and enabling it does not grant trusted access.
• If the check cannot be completed — access denied, or an ambiguous response — you MUST treat the result as inconclusive, not negative. Report that and stop. You MUST NOT enable trusted access on that basis.
• If the enable call is denied, you MUST report that the caller is not in the management account and stop. A delegated administrator can read the setting but cannot change it.

Step 3 — Create the domain access role

This is the role the service assumes to operate on Spaces across the organization. It MUST live in the management account, and unlike the account-scoped Space role it is verified at create time — the service performs a real sts:AssumeRole against it while handling create-domain-for-organization, and the create fails if that does not succeed.

The ARN must match arn: followed by the partition, then :iam::<management-account-id>:role/<role-name>, and be 20–2048 characters.

Create the role

Call create-role with AssumeRolePolicyDocument set to this trust policy. The aws:SourceAccount and aws:SourceArn conditions are confused-deputy protection — keep both. Note the resource type is organization-domain, not domain.

Attach the managed policy

Constraints:

• The role MUST be in the management account. A role in any other account is rejected, and the message names the management account requirement.
• The aws:SourceArn pattern MUST use the organization-domain resource type and MUST stay wildcarded across Region and domain ID. The domain ID does not exist yet when the role is created, so a narrower pattern cannot match.
• You MUST NOT substitute a *FullAccess policy or hand-write a broad inline policy. The Domain’s permissions are defined by the managed policy above.
• You MUST NOT skip creating the role and hope create-domain-for-organization provisions one. The role ARN is a required input.
• If the create-time assume check fails, you MUST report the trust policy as the cause and stop. Do NOT retry the Domain create — it will fail identically until the trust policy is fixed.
• You MUST NOT attach the managed policy to a role the customer supplied without asking. They may be reusing it, and widening someone else’s role is not yours to do.

Step 4 — Create the Domain

Check what already exists first

list-domains reports the Domains visible to the caller, of both scopes, and each entry carries its ARN. Use the ARN’s resource type to tell them apart — organization-domain for an org Domain, domain for an account-scoped one. This is the check to run when the customer has no domain ID yet.

Read the result before creating anything:

• An organization-domain entry exists — the org Domain is already set up. Report its name and ID from the summary and its endpoint URL from get-domain-for-organization (summaries do not carry it), confirm the customer wants to use it, and continue from Step 5 with its domainId.
• Only a domain entry exists — an account-scoped Domain occupies the management account, and it blocks the org Domain. Tell the customer it must be deleted before an org Domain can be created, and that deleting it is their decision. Then stop. Do not create anything, and do not describe the situation merely as blocking.
• Nothing comes back — proceed.

Constraints:

• You MUST run this check before create-domain-for-organization.
• If an account-scoped Domain occupies the management account, you MUST say plainly that it has to be deleted before an org Domain can be created, and that deleting it is the customer’s decision. Do NOT describe it vaguely as blocking.
• You MUST distinguish the two ARN resource types rather than treating any returned Domain as the org Domain. Acting on an account-scoped Domain as though it were org-scoped produces confusing failures several steps later.
• If the check cannot be completed — access denied, or an ambiguous response — you MUST treat the result as inconclusive, not negative. Report that and stop. You MUST NOT create a Domain on that basis.

Create it

For an IAM-only Domain:

For an Identity Center Domain, identityProviderConfiguration becomes required and carries the instance ARN:

Capture domainId, domainArn, domainEndpointUrl, organizationId, and ownerAccountId from the organizationDomain object in the response (the same object get-domain-for-organization and update-domain-for-organization return).

Constraints:

• You MUST create an Identity Center Domain in the instance’s primary Region. The service verifies this; the rejection names the primary Region and the current Region, so read it rather than guessing.
• You MUST capture domainEndpointUrl and give it to the customer. Every member account reaches Omni through it.

Step 5 — Create the Spaces

A Space lives in a member account. There are two ways to create one, and the choice was made in Step 1.

The Space itself — its name, its space access role, its encryption — works exactly as it does under an account-scoped Domain. See references/cloudwatch-omni/spaces-and-domains.md, picking up at its Space steps. The space access role is still required and still belongs in the account that will own the Space.

Path A — the member account creates its own Space

The member account calls create-space directly, passing the shared domainId. Nothing about the call differs from the account-scoped path except that the Domain was not created by that account.

Path B — the management account creates it on the member’s behalf

The management account, or a registered delegated administrator, vends credentials for the target account and uses them to create the Space:

--credential-type is required, and SPACE_OPERATION is its only value — always pass it, and do not treat it as a choice to put to the customer. Omitting it is rejected by the client before the request is sent.

The response carries temporary credentials for the target account. Use them to call create-space against that account. Once a Space exists, credentials can be vended for it directly by passing spaceId in the context instead of the domainId/targetAccountId pair.

Enumerate what exists across the organization at any point:

Constraints:

• You MUST NOT tell a member account to call create-domain first. The org Domain already exists and a second Domain in the member account is neither needed nor permitted alongside it.
• On Path B you MUST use the vended credentials for the create-space call itself. Calling create-space with the management account’s own credentials creates a Space in the management account, not the target account.
• Region availability is narrower than it looks. A Space may live in a Region other than the Domain’s, but not every Region is available, and under an Identity Center Domain the Region must also be one Identity Center is replicated to.

Step 6 — Grant Domain access

A Domain with no grants has no administrators. Grant the principals the customer named in Step 1.

A grant applies to the entire Domain. It conveys administration of every Space in the Domain, including Spaces created later. There is no way to scope a Domain-level grant to one account or one Space — that is what per-Space grants are for.

name is required: 1–64 characters matching ^[a-zA-Z0-9_-]+$. permission takes the single value ADMIN. principalType is one of IDC_USER, IDC_GROUP, IAM_USER, IAM_ROLE, or IAM_ROOT.

Capture the grant ID from the accessGrant object; it is what delete-domain-access-grant-for-organization takes if a single grant needs revoking later.

Review the grants with:

Constraints:

• permission is a single value and it must be ADMIN. The account-scoped permission values — READ, READ_WRITE_DELETE, SPACE_ADMIN, and CUSTOM — are not valid at the Domain level.
• You MUST state the blast radius when confirming a grant: the principal will administer every Space in the organization’s Domain.
• For an Identity Center Domain you SHOULD prefer IDC_GROUP over IDC_USER. Group membership changes without touching grants.
• You MUST NOT grant IAM_ROOT unless the customer explicitly asks. It is the broadest principal available and rarely what they mean.

Verifying the setup

Read the Domain back rather than trusting the create response, and confirm the Spaces and grants landed:

Then report to the customer, in one line: the domain endpoint URL, the domain ID, the organization ID, how many Spaces exist, and how many administrators were granted.

Constraints:

• You MUST confirm at least one ADMIN grant exists before declaring setup complete. A Domain nobody administers is not a working setup.

Cleanup

Order matters, and it spans accounts.

1. Every Space is deleted. The Domain cannot be deleted while any Space is associated with it. Use list-spaces-for-organization to enumerate what remains. A Space is deleted by its owning account, or by the management account using credentials vended through get-space-credentials-for-organization.

2. Delete the Domain, from its home Region — the Region it was created in:

   Copy

   aws___call_aws → aws cloudwatchomni delete-domain-for-organization --domain-id <domain-id>

   Use delete-domain-for-organization, not delete-domain — the account-scoped operation rejects an org Domain and tells you to use this one instead.

   Deleting the Domain removes its access grants. Do not revoke them first as hygiene; it is unnecessary work. delete-domain-access-grant-for-organization is for revoking one principal’s access while the Domain stays in use.

3. The domain access role — branch on who created it:

   • The agent created it in Step 3: detach the managed policy with detach-role-policy, then delete-role. A role with policies attached cannot be deleted.
   • The customer supplied it: do NOT delete or modify it.

4. Trusted access can be left enabled. If the customer wants it off, that is an organization-wide change made through CloudWatch’s own telemetry configuration controls rather than by disabling the Organizations integration directly.

Constraints:

• You MUST delete every Space before the Domain, and you MUST say plainly which accounts are involved. Spaces in member accounts need either that account’s credentials or credentials vended for it.
• You MUST delete the Domain from its home Region. A delete from anywhere else is rejected and the message names the home Region.
• You MUST use delete-domain-for-organization. delete-domain rejects an org Domain and tells you to use this operation instead.
• Deleting Spaces is destructive. You MUST confirm with the customer first and say plainly that each Space’s telemetry goes with it.
• You SHOULD NOT disable trusted access as part of cleaning up a Domain. It is organization-wide and other CloudWatch features depend on it.

Troubleshooting

Rule: When a call returns an error, surface the error code and message verbatim, then map to the mitigation below. Do NOT invent error text, do NOT paraphrase what the service returned, and do NOT synthesize a mitigation for an error that is not listed.

If the error does not match a row above, quote it verbatim, say it is unmapped, and ask the customer how to proceed.

Expectations that look like faults

• Every org operation fails authorization at once. That is trusted access being off, not a permissions problem on the caller.
• A Domain-level grant cannot be made read-only, or scoped to one account. ADMIN over the whole Domain is the only shape. Narrower access is granted per Space.
• A Space cannot be created in some Region. Region availability under an org Domain is narrower than it appears, and under an Identity Center Domain the Region must also have Identity Center replicated. Nothing about the Domain or the Space request is wrong.
• Grants disappear when the Domain is deleted. That is expected — the service removes them with the Domain.

Security considerations

• The domain access role is assumed for callers across the whole organization. It is the broadest role in this setup — keep both aws:SourceAccount and aws:SourceArn conditions on its trust policy, keep the organization-domain resource type pinned, and keep the role out of any workflow that does not need it.
• Keep the domain access role in the management account and nowhere else. The service requires this, and it keeps organization-wide capability from sitting in a member account.
• A Domain access grant reaches every Space in the Domain. Treat it as an organization-wide grant, not an account-level one, and keep the list short.
• Prefer IDC_GROUP over IDC_USER for Domain administrators so that offboarding a person does not require finding and revoking their grant.
• Avoid IAM_ROOT as a grant principal. Root has no session identity to attribute actions to.
• Credentials from get-space-credentials-for-organization are temporary and act in another account. Do not persist or log them, and do not reuse them beyond the operation they were vended for.
• Enumerate grants with list-domain-access-grants-for-organization periodically.
• Delegated administrator is a standing organization-wide capability. Register only accounts the customer intends to keep in that role.
• Trusted access is organization-wide and shared with other CloudWatch features. Enabling it is a deliberate decision for the organization administrator, not a side effect of setting up one Domain.

Additional resources

• references/cloudwatch-omni/spaces-and-domains.md — account-scoped Domains, and the Space steps used under an org Domain
• references/cloudwatch-omni/instrumentation/collector.md — deploying a collector that exports to CloudWatch’s OTLP endpoints, for telemetry not yet reaching CloudWatch
• references/cloudwatch-omni/data-forwarding-and-centralization.md — forwarding telemetry that is already in CloudWatch log groups into the Dataset