See every HPE server the way HPE Compute Ops Management sees it, right inside HPE Morpheus. Health, firmware, hardware, warranty and alerts for each physical host, a status rollup for each cluster, and a few safe one-click fixes. No console switching, no agents, no iLO passwords.
Unofficial community project. Not an HPE or Morpheus Data product. Not developed, endorsed or supported by Hewlett Packard Enterprise. Provided as is, without warranty. HPE, HPE GreenLake, Compute Ops Management, HPE Morpheus, HPE Morpheus VM Essentials, ProLiant and iLO are trademarks of Hewlett Packard Enterprise; see NOTICE.md for trademarks and for every source this project was built from.
A real Morpheus 9.0.1 appliance with three ProLiant servers in COM. ① The plugin's HPE COM tab on the cluster page. ② Totals for the whole cluster. ③ Every host, its COM health, power, alerts, support contract and COM maintenance state, with a link to the server in COM. ④ Straight from HPE Compute Ops Management, at most a minute old. Names and serial numbers are example values.
- One pane of glass. The people who run your VMs see what COM knows about the hardware underneath, on the same page: health, iLO security, firmware, processors, memory, drives, warranty and active alerts.
- The whole cluster at a glance. One table answers "is any host in this cluster unhealthy, alerting, out of support or in maintenance?", with a link to each host and to its COM page.
- Nothing to install on your hosts. The plugin matches each Morpheus host to its COM server using the network-adapter MAC addresses both systems already know. No agents, no SSH, no IPMI, no iLO credentials in Morpheus.
- Safe, one-click fixes. Light the UID LED to find the box, refresh COM's inventory, reset a stuck iLO, or put a server into COM maintenance before hardware work. Every button restores, observes or refreshes; none reconfigures. Why these actions?
- Secure by design. The API secret never leaves the Morpheus appliance. Buttons are off by default and gated by a Morpheus role permission. Every action request passes a chain of server-side checks. Two independent security reviews have been done, and every finding is fixed. Security details
- Drop-in. One jar for Morpheus 9.0 and 9.1, with a settings form, two role permissions, and a test suite that runs against both plugin APIs. 0.5.0 is tested on a real Morpheus 9.0.1 appliance against the live COM API.
- Keeps working when COM changes. When the live COM API drifted from its published reference in October 2026, 0.4.2 adapted the same day: smaller pages, fallback reads and the newer GreenLake sign-in. See Known issues.
- Download
com-morpheus-plugin-<version>.jarfrom the latest release. - In HPE GreenLake, create an API client for Compute Ops Management.
- In Morpheus, upload the jar (Administration → Integrations → Plugins), enter the client ID and secret, and grant the com-morpheus permissions to the right roles.
- Open a physical HPE host or a cluster: the HPE COM tab is there.
Step-by-step instructions: docs/INSTALL.md. How the pieces fit together: How it works.
These screens are from a real Morpheus 9.0.1 appliance. Names, serial numbers and addresses are replaced with example values (host-a, XYZ0000001, 192.0.2.21), the same ones the install guide uses, so you can follow one server from screen to screen.
- The HPE COM tab sits beside Morpheus's own tabs on every physical HPE hypervisor host.
- Status at a glance: COM health, iLO security, active alerts, power, and how the host was matched to its COM server (here, every MAC address agrees).
- One-click actions (optional, off by default): UID light, Collect inventory, iLO reset and COM maintenance, run as COM jobs with live progress.
- Seven sub-tabs mirror the server's page in COM: Summary, Health, Firmware, Hardware, Storage, Support and Alerts.
- Open in COM and Open iLO take you straight to the server in COM or to its iLO.
- The server as COM sees it: serial, iLO, firmware baseline, OS, subscription and more, without leaving Morpheus.
The cluster tab at the top of this page answers "is anything wrong in this cluster?" in one look. When something needs attention, it says so. Here's a larger example, rendered from the plugin's test data: a warning, active alerts, an expired support contract, a server in COM maintenance and a host COM doesn't know.
Morpheus puts extra tabs under the ▼ at the right end of the tab row. The plugin's tab is called HPE COM:
Rendered from the plugin's test data, with the same example values as above.
Host tab. The COM server page, mirrored in seven sub-tabs, with links to COM and the iLO and optional action buttons:
COM maintenance mode. Mute COM's support cases and notifications for a server during planned work. The tab makes it obvious while it's on:
| Health | Hardware | Storage |
|---|---|---|
![]() |
![]() |
![]() |
| Firmware | Support | Alerts |
|---|---|---|
![]() |
![]() |
![]() |
- Appears only where it makes sense: on hypervisor hosts whose hardware vendor is HPE and that aren't virtual machines.
- Finds the matching COM server automatically. Morpheus already stores each host's factory network-adapter MAC addresses, and COM's inventory lists the same adapters. The plugin matches them, so it never needs access to the host itself. For special cases, a
com-serial=<serial>orcom-id=<COM id>tag on the host overrides the match. - Mirrors the COM server page in seven sub-tabs:
- Summary: state, model, serial, iLO, baseline, power, subscription and more.
- Health: every component, plus iLO security settings with COM's recommended actions.
- Firmware: every component and version.
- Hardware: processors, memory, network adapters and power supplies.
- Storage: volumes and drives.
- Support: warranty and contract.
- Alerts: active alerts.
- Links:
- Open in COM goes to the server's page in COM.
- Open iLO goes to the iLO's address, as COM's own Remote console link does.
- Cluster summary (new in 0.4.0): an HPE COM tab on each cluster's page lists every host with its COM match, model, serial, health, power, active alerts, support state and COM maintenance mode, with totals and links to each host and its COM page. It's read-only. It reads the cluster's own COM servers (by id) and the warranties once for the whole cluster, and active alerts in one workspace-wide query where COM allows it, otherwise per host.
- Optional action buttons: UID light on/off, Collect inventory, iLO reset, and Enter/Exit COM maintenance (new in 0.4.0), run as COM jobs with live progress. See Why these actions and Security.
The action set is small on purpose. Each button restores, observes or refreshes. None of them reconfigures. Nothing here changes BIOS, iLO or storage settings, flashes firmware, or touches the host's power. That keeps the riskiest class of remote change, configuration drift you didn't intend, out of a one-click UI. You still get direct, auditable control of the resources most likely to need a nudge.
The iLO is the server's independent management brain. It sits beside the host and is used for monitoring, remote console, Redfish, and the link to Compute Ops Management. Like any embedded controller, it can occasionally become sluggish or unresponsive: the web UI stops loading, remote console sessions hang, or COM reports the server as disconnected. A reset is the controller's equivalent of "turn it off and on again", and it's almost ideal for a one-click action:
- It changes nothing, it restores. The iLO restarts with exactly the settings, users, certificates and licenses it had. No configuration is written, so there's nothing to review, roll back or drift.
- The workload is left alone. Only the management controller restarts. The hypervisor host keeps running, and so do its VMs and storage. The only interruption is to management, for a minute or two: the iLO web UI, open remote console sessions and COM's connection. All of them come back on their own.
- It's a small, predictable blast radius. One server's controller, one well-understood outcome, and repeating it is harmless.
- It goes through the front door. The reset is a COM job under the same governed identity as everything else. It's recorded in COM, and in Morpheus with the user who asked for it. It needs no iLO credentials in Morpheus and no network path from a browser to the iLO.
- It shortens the path to recovery. The person looking at a misbehaving host in Morpheus can bring its management plane back without switching consoles or raising a ticket. For anything deeper, Open in COM and Open iLO are one click away.
It still asks for confirmation first, because a reset briefly drops the management connection.
- UID light on/off turns the blue locator LED on the chassis on or off. It's the safest possible action, and it answers the most practical data-center question: which box is it?
- Collect inventory asks COM to refresh what it knows about the server. Hardware, storage and network adapters are re-read, so the Hardware and Storage sub-tabs and the MAC matching reflect reality after a change, such as a replaced DIMM or NIC.
Enter COM maintenance and Exit COM maintenance switch the server's maintenance mode in COM. While it's on, COM suspends automatic support case creation, ServiceNow incidents and email notifications for that server (HPE developer documentation). Use it before planned hardware work, so a pulled drive or a power-off doesn't open a support case.
- It changes COM's bookkeeping, not the server. Nothing is written to the server, its iLO, the host or its VMs.
- It's separate from Morpheus maintenance mode on purpose. The plugin never turns one on because of the other. Putting a host into Morpheus maintenance moves workloads; COM maintenance only mutes COM's notifications. Use whichever you need, or both.
- Both buttons ask to confirm, because while COM maintenance is on, real faults on that server won't raise cases or notifications. The tab shows which state the server is in, and offers only the opposite action.
Power on, power off, restart, firmware updates and configuration changes are not in this version. They can affect the host and its virtual machines, and they deserve stronger safeguards than a button. For example, power actions should be allowed only when the Morpheus host is in maintenance mode with no running VMs. Firmware updates are better done in COM itself, where baselines and reboot policies are managed.
This selection reflects the current version and may change as the plugin gains further functions.
- HPE Morpheus 9.0.x or 9.1.x, including HPE Morpheus VM Essentials.
- Physical HPE servers that Morpheus manages as hypervisor hosts, such as HVM (HPE Virtual Machine) hosts.
- HPE servers onboarded in an HPE GreenLake workspace with Compute Ops Management.
- A GreenLake API client for Compute Ops Management. Ideally it belongs to a dedicated GreenLake account, because the plugin acts as that account (see Security).
- A view-only COM role is enough for the tab.
- The action buttons need a COM role that can run server jobs.
- Outbound HTTPS from the Morpheus appliance to
<region>.api.greenlake.hpe.com,global.api.greenlake.hpe.comandsso.common.cloud.hpe.com. A proxy is supported.
Use the prebuilt jar from the Releases page. There's no need to build anything.
One jar, com-morpheus-plugin-<version>.jar, works on Morpheus 9.0.x and 9.1.x. It's compiled against the oldest supported Plugin API (1.4.1, Morpheus 9.0), and its whole test suite also runs against Plugin API 1.5.0 (Morpheus 9.1). Before each release, tools/check-plugin-core.sh checks HPE's published plugin sources and docs for any compatibility change.
Each release includes SHA256SUMS. Check the download with sha256sum -c SHA256SUMS.
Then follow docs/INSTALL.md: create the COM API client, upload the jar, enter the settings, and grant the permissions.
To report a vulnerability, see SECURITY.md.
The plugin uses one GreenLake API client for the whole Morpheus appliance: every tenant, host and user. It's set once in the plugin settings. A GreenLake personal API client inherits the roles of the GreenLake user who created it. So:
- Everything the plugin does in COM runs as that GreenLake user. That covers reading data and every action button, whichever Morpheus user clicked. COM's activity log shows that identity, not the Morpheus user.
- The client can reach whatever that user's COM role allows, usually the whole workspace. The Morpheus tenant check below limits which hosts a Morpheus user can pick. It doesn't limit what the client could do in COM.
- The Morpheus permission is the only per-user control. Who may press the buttons is decided entirely by com-morpheus: Run HPE COM actions in Morpheus. Morpheus's log is the only record of which Morpheus user started each job.
- The client depends on its owner. If that GreenLake user leaves or loses access, the plugin stops working. Until then, their name is on every action in COM.
Therefore:
- Create the API client under a dedicated GreenLake account (a service or shared admin account under your organization's rules), not a person's own account.
- Give that account the smallest COM role that works: view-only while Enable actions is off; an operator role only if you turn actions on. Limit it to the relevant servers if your workspace supports scoped access.
- Grant Run HPE COM actions only to roles you would trust with that account.
- Keep Morpheus's log, which ties each job to a Morpheus user.
- Credentials stay on the appliance and go only to HPE. Only the Morpheus appliance talks to GreenLake. The API client secret is stored with the plugin settings and is never sent to browsers. It's sent only to HPE GreenLake sign-in addresses: the URL settings must be under
greenlake.hpe.comorcloud.hpe.com, so someone who can edit the plugin settings can't point them anywhere else. - Read-only by default. The action buttons are off until an administrator turns on Enable actions. Even then, they appear only for roles granted com-morpheus: Run HPE COM actions at full access.
- Every action request is re-checked on the appliance (hardened in 0.3.1 after an independent security review):
- Request:
- it must be a POST (anything else, including an unknown method, is refused);
- it must carry the
X-Requested-Withheader, which a cross-site form or image can't send; - when the browser names its origin, that origin must be this appliance;
- Morpheus's own CSRF protection may apply as well; the plugin doesn't rely on it.
- The tab's action token: the request must carry the signed token the server put into that host's tab. Markup injected elsewhere in the UI can't forge a working button. Tokens expire after a day.
- Permission and setting: actions must be turned on, and the user must have Run HPE COM actions at full access.
- Ownership: the user's tenant must own the host (or the user is in the master tenant), and the host must be a hypervisor host.
- Which server: the host's COM server is found again from Morpheus data, never from the browser.
- Only confirmed matches qualify: all of the host's MAC addresses match one COM server.
- A
com-serial/com-idtag only changes which server is shown. For actions, a tag must agree with the host's own hardware: at least one of its MAC addresses must be in that server's inventory, and none in another server's. So a user who can edit tags can't aim a job at someone else's server. A tag still decides which server's data the tab displays. Because everything is read through the one shared GreenLake client, a user who can edit a host's tags can view (not act on) any server in the COM workspace, on the host tab and in the cluster summary. Limit the client's COM scope if that matters to you.
- Cluster summary: it lists only hosts owned by the cluster's tenant, and only users of that tenant (or the master tenant) can open it.
- Results: job progress is shown only for that server's own jobs. Unknown, foreign and malformed jobs all get the same answer.
- Rate: one action per server every 30 seconds, and one iLO reset per server every 10 minutes. Each Morpheus user can make 60 requests a minute to the action routes and start 20 actions an hour, so one user can't use up the shared client's COM rate limit.
- Errors: unexpected errors return a reference number; the details stay in the Morpheus log.
- Request:
- No commands on hosts. The plugin never runs anything on hypervisor hosts. Actions are COM jobs only, and each one is logged with the Morpheus user who started it.
- No power actions. Power on, off and restart aren't in this version.
Before 0.3.1, the action routes, permission model and browser script were reviewed independently, by a separate AI reviewer (Anthropic's Claude Fable model) given read-only access to the source and HPE's published plugin API source. It found:
- one High: a
com-id/com-serialtag could aim an action at any COM server the API client could reach, beyond the user's tenant; - two Medium: cross-site request protection relied entirely on Morpheus, and the method check failed open; markup injected elsewhere in the UI could trigger the page-wide script;
- several Low/Informational findings: job-status checks, error detail, no rate limit, a missing host-type re-check, and the
/capspermission.
All of them are fixed in 0.3.1, each with tests, and the fixes were re-reviewed. Versions 0.2.0 and 0.3.0 are affected when Enable actions is on: upgrade to 0.3.1. Exploiting any of them required a signed-in Morpheus user with the actions permission; none was reachable anonymously.
Before 0.4.0, the same reviewer checked the new cluster summary and COM maintenance actions. It found two Medium issues (the cluster tab didn't check tenants, and it read the workspace's whole alert history on each refresh) and two Low ones (COM maintenance needed more safeguards, and the tab appeared on clusters without HPE hosts). All were fixed before release, with tests, and no released version was affected.
On October 31, 2026, the same reviewer did a full review of the code, scripts, build, docs and public history. It found no Critical issue, and no way for a user without the right Morpheus permission to run an action or read another tenant's host data. It found:
- one High:
tools/create-com-user.py(new in 0.4.2) gave its role Read on far more of Morpheus than the tabs need, including logs, tasks and templates; - three Medium: the URL settings weren't limited to HPE, so someone who can edit the plugin settings could send the client secret elsewhere; the response cache grew with the size of the COM workspace; and there was no per-user limit on the action routes;
- several Low/Informational findings: iLO link checks, error detail, token lifetime, an unused route, group lookups during page loads, and build hygiene.
The High, the Mediums and most of the Lows are fixed in 0.5.0, with tests. CI with signed build provenance is planned separately. If you made a role with the 0.4.2 script, delete it and run the new script.
In short:
- Only the Morpheus appliance talks to GreenLake, over outbound HTTPS with one OAuth2 API client. The browser gets HTML rendered on the appliance, and sends action requests back to it.
- Matching uses data Morpheus and COM already have. A host's stored network-interface MACs are compared with the adapter MACs in COM's inventory. Nothing runs on the host.
- Reads are cached (1 minute for server state and alerts, up to 24 hours for firmware bundles), so opening tabs doesn't hammer the COM API.
- Actions pass a chain of checks on the appliance before a COM job is created. See Safeguards in the plugin.
The infographic is generated by tools/draw-how-it-works.py; edit the text there and run it again to update it.
The source is commented throughout: ComService (matching and actions), ComClient (COM API), ComController (routes), ComView / ComInventoryView (what the tab shows).
You need only Docker; the build runs in a JDK 17 container.
tools/build.shThis builds and tests the jar into dist/, then repeats the test suite against the newer Plugin API (1.5.0). A first build downloads Gradle and its dependencies; later builds take a few minutes.
Optional checks (opt-in, off by default):
COM_MOCK=1runs the real HTTP client against HPE's public COM mock server (LiveChecksSpec).PREVIEW_OUT=<dir>renders the tab to HTML for review (LiveChecksSpec).LOAD_JARS=<jar>[,<jar>…]loads built jars through the plugin API's ownPluginManager.registerPlugin(PluginLoadSpec). Use a renamed build, because the loader is parent-first.
The live COM API has drifted from its published reference since October 2026. Each of these is handled from 0.4.2 on.
- COM caps every page at 500 items (HTTP 400,
limit"less than or equal to 500"), although its API reference allows 1000 for alerts. This was the cause of the 0.4.0 cluster-summary alerts error ("COM /v1/alerts: HTTP 400", with a count of 0 for every host) and, from October 29, 2026, of failed alert reads on host tabs in 0.4.1. 0.4.2 never asks for more than 500. - COM refuses
select=serverId,device/idon alerts ("Cannot select unknown properties"), although its reference lists both fields. 0.4.2 doesn't useselectfor alerts and reads the server id from the full alert. GET /v1/servers/{id}can answer HTTP 500 for servers that COM's own console shows as healthy (seen October 29, 2026). The host tab then showed "Couldn't load HPE Compute Ops Management data". 0.4.2 tries COM's other published path (/v1beta2), then shows what the server list has, with a note, so the tab, its sub-tabs and the action buttons still work.- Legacy-SSO tokens: some HPE GreenLake services started refusing tokens from the legacy sign-in on October 29, 2026. 0.4.2 signs in through the workspace issuer first when the workspace ID is set, and moves on to the next sign-in URL if COM refuses a token. Set the workspace ID in the plugin settings.
- If COM refuses the workspace-wide alerts query for any other reason, the cluster summary reads each matched server's alerts instead (since 0.4.1), and alert counts stay correct.
- One COM workspace and region per Morpheus appliance.
- The COM web UI address is confirmed for US West (
us-west2.compute.cloud.hpe.com). EU Central and AP Northeast follow the same pattern but are unverified; the COM web UI URL setting overrides it. - Supported: Morpheus 9.0.x and 9.1.x. A new Morpheus release with a new Plugin API is checked with
tools/check-plugin-core.shand the compatibility test run before it's listed as supported.
Pull requests are welcome. External development is how plugins like this get better. See CONTRIBUTING.md for how to build, test and submit changes. Use the issue templates for bugs and ideas, and report security problems privately.
This plugin was written from scratch, using HPE's public documentation and the Morpheus plugin API as references. NOTICE.md lists every source, its license and exactly how it was used, plus the trademark notices.
MIT for this project's own code. The Morpheus plugin API it compiles against is licensed separately by Morpheus Data, LLC and isn't included.











