<iframe src="https://www.googletagmanager.com/ns.html?id=GTM-M74D8PB" height="0" width="0" style="display:none;visibility:hidden">
Loading
Skip to NavigationSkip to Main Content
0D5WR00002IYlva0ADOkta Classic EngineLifecycle ManagementAnswered2026-10-30T18:06:32.000Z2026-10-19T06:57:32.000Z2026-10-30T18:06:32.000Z

seoheel.92373 (Customer) asked a question.

Inquiry Regarding OPA/OPS VM Deployment and High Availability (HA) Support

Hello,

 

We are currently reviewing the VM server architecture for deploying the Okta Provisioning Agent (OPA) and Okta On-prem SCIM Server (OPS) in a customer environment.

The customer has asked whether a two-VM high availability (HA) configuration is mandatory and whether VMware virtualization HA features can be used as an alternative.

We would like to confirm the officially supported deployment options and recommended HA architecture.

 

[Planned Deployment Environment]

  • Components: OPA + OPS (Traditional On-premises Provisioning architecture)
  • Deployment Option: Installing both OPA and OPS on a single VM server
  • HA Options Under Consideration: Deploying two VM servers for redundancy or utilizing VMware virtualization HA features

 

[Findings from Okta Official Documentation]

According to Okta's official documentation, the HA configuration requires installing additional OPA and SCIM Connector components on separate servers. The system requirements documentation also specifies a dedicated Linux server for OPS installation.

However, we could not clearly determine whether the proposed single-VM deployment is officially supported or whether VMware virtualization HA features can replace application-level redundancy.

We would appreciate your clarification on the following questions.

 

[Questions]

1. Single-VM Deployment and HA Requirements

  • Is it officially supported to install and operate both OPA and OPS on the same VM server?
  • Is a two-VM redundant configuration mandatory for all deployments, or is it required only when implementing high availability?

 

2. VMware Virtualization HA as an Alternative

  • Is a deployment consisting of OPA and OPS on a single VM, protected by VMware HA, officially supported by Okta?
  • Can VMware virtualization features, such as VMware HA, vMotion, and Live Migration, replace a separate application-level HA configuration for OPA/OPS?

 

3. Application-Level High Availability Requirements

  • Is the HA architecture described in Okta's documentation intended to protect against both physical server failures and OPA/OPS application-level failures?
  • If application-level redundancy is required, what is the officially supported and recommended server architecture?
  • Do the HA requirements described in the official documentation for SCIM Connector-based deployments also apply to the OPA + OPS configuration currently under consideration?

 

If possible, please provide confirmation of the officially supported configurations, recommended architecture, and relevant Okta documentation for each question.

 

[References – Okta Official Documentation]

  1. On-premises Provisioning Architecture
  2. https://help.okta.com/en-us/content/topics/provisioning/opp/opp-architecture.htm
  3. System Requirements for On-premises Connector – Generic Databases
  4. https://help.okta.com/en-us/content/topics/provisioning/opc/on-prem-generic-db-sys-req.htm

 

Thank you for your assistance.


  • Paul S. (Okta, Inc.)

    Hello @seoheel.92373 (Customer)​ Separate servers isolate the two components' failures, resources, lifecycles, and network paths, which makes the deployment easier to operate, scale, and support.

    OPA and OPS have different roles. OPA relays provisioning requests from Okta's cloud to the on-premises network. OPS is the SCIM endpoint that receives those requests and writes to your target system. Each role has its own network, resource, and maintenance needs, so combining them on one host creates shared points of failure.

     

    Use one server for OPA and a separate server for OPS, following the separate-server topology described in the On-premises Provisioning Architecture documentation. The main technical reasons are:

    1. Isolate failure domains. A crash, memory exhaustion, or runaway process in OPS (for example, a heavy import job) does not stop OPA's connection to Okta, and an OPA problem does not take down the SCIM endpoint.
    2. Separate resource consumption. OPS handles bulk SCIM traffic and target-system queries, while OPA runs a persistent outbound connection. Separate hosts prevent one workload from starving the other of CPU, memory, or file handles.
    3. Align network placement with trust boundaries. OPA needs outbound HTTPS access to Okta. OPS needs inbound access from OPA and outbound access to the target application or database. Separate hosts let you apply different firewall rules, allowlists, and network zones to each component.
    4. Enable independent maintenance. You can patch, upgrade, or restart OPS without interrupting the OPA service, and the reverse. This also keeps the OPA and OPS version and dependency changes separate.
    5. Support clear troubleshooting. Logs, process states, and resource metrics map to one role per host, so the source of an error is easier to identify when you work with Okta Support.
    6. Scale each tier independently. High availability requires a second OPA and a second SCIM connector instance. Separate hosts let you add capacity to the OPS tier without changing the OPA tier.

     

    Expand Post
    Selected as Best
  • Paul S. (Okta, Inc.)

    Hello @seoheel.92373 (Customer)​ Thank you for posting on our Community page!

     

    You are asking whether Okta Provisioning Agent (OPA) and Okta On-Premises SCIM Server (OPS) can both run on one virtual machine, whether a two-VM high availability (HA) configuration is mandatory, and whether VMware HA can replace application-level redundancy for these components.

     

    Summary: Okta documents OPA and the SCIM connector (OPS) on separate servers, with application-level HA built from additional agent and connector instances on separate servers. VMware HA protects against host failure only and does not replace that application-level redundancy.

     

    Solution: Use separate servers for OPA and OPS in production, and deploy the second OPA and SCIM connector instances on a second server for HA. Treat VMware HA as an infrastructure layer on top of this design, not as a substitute for it.

    1. Review the On-premises Provisioning Architecture documentation to confirm the separate-server topology and the component placement rules: https://support.okta.com/help/s/question/0D54z00008glHEMCA2 (Community reference to the architecture page: https://help.okta.com/en-us/Content/Topics/Provisioning/opp/OPP-architecture.htm).
    2. Deploy OPA on one server and OPS on a dedicated Linux server for the primary environment. Check the generic database system requirements page for OPS server sizing and OS support.
    3. Deploy a second OPA and a second SCIM connector instance on a separate server to meet the documented HA pattern. Two VMs are required when application-level HA is required.
    4. Apply VMware HA, vMotion, or Live Migration only as an infrastructure safeguard. These features restart or move a VM after a host failure but do not detect or recover from an OPA or OPS process failure, so they do not satisfy the application-level HA requirement.
    5. Open a case with Okta Support if the customer must co-locate OPA and OPS on one VM. Okta Support can confirm whether that topology is supported for the customer's tenant and version. This confirmation should be obtained in writing before production rollout.

     

     

    Thank you for reaching out to our Community and have a great day!

    --

    Help others in the community by liking or hitting Select as Best if this response helped you.

    Expand Post
  • seoheel.92373 (Customer)

    Thank you for your detailed response.

     

    I have reviewed your explanation and would like to ask one additional question.

    You recommended installing the Okta Provisioning Agent (OPA) and On-Premises SCIM Server (OPS) on separate servers.

    Could you please clarify the specific technical reasons for this recommendation?

     

    Thank you for your continued support.

    Expand Post
    • Paul S. (Okta, Inc.)

      Hello @seoheel.92373 (Customer)​ Separate servers isolate the two components' failures, resources, lifecycles, and network paths, which makes the deployment easier to operate, scale, and support.

      OPA and OPS have different roles. OPA relays provisioning requests from Okta's cloud to the on-premises network. OPS is the SCIM endpoint that receives those requests and writes to your target system. Each role has its own network, resource, and maintenance needs, so combining them on one host creates shared points of failure.

       

      Use one server for OPA and a separate server for OPS, following the separate-server topology described in the On-premises Provisioning Architecture documentation. The main technical reasons are:

      1. Isolate failure domains. A crash, memory exhaustion, or runaway process in OPS (for example, a heavy import job) does not stop OPA's connection to Okta, and an OPA problem does not take down the SCIM endpoint.
      2. Separate resource consumption. OPS handles bulk SCIM traffic and target-system queries, while OPA runs a persistent outbound connection. Separate hosts prevent one workload from starving the other of CPU, memory, or file handles.
      3. Align network placement with trust boundaries. OPA needs outbound HTTPS access to Okta. OPS needs inbound access from OPA and outbound access to the target application or database. Separate hosts let you apply different firewall rules, allowlists, and network zones to each component.
      4. Enable independent maintenance. You can patch, upgrade, or restart OPS without interrupting the OPA service, and the reverse. This also keeps the OPA and OPS version and dependency changes separate.
      5. Support clear troubleshooting. Logs, process states, and resource metrics map to one role per host, so the source of an error is easier to identify when you work with Okta Support.
      6. Scale each tier independently. High availability requires a second OPA and a second SCIM connector instance. Separate hosts let you add capacity to the OPS tier without changing the OPA tier.

       

      Expand Post
      Selected as Best

Loading
Inquiry Regarding OPA/OPS VM Deployment and High Availability (HA) Support
Okta Support Agent (Beta)

Terms of Service

Hello! I'm the Okta Support Agent (Beta), your go-to resource for questions about Okta products. Please remember for your security and privacy, do not share any Personally Identifiable Information (PII) such as passwords, IP addresses, or other details with me.

Okta Privacy Policy and Terms of Service apply