Skip to content

com-morpheus-plugin 0.5.0

Latest

Choose a tag to compare

@grahamfairbank grahamfairbank released this 31 Oct 18:02

Tested on Morpheus 9.0.1 with the live COM API. This release has two parts: security hardening and faster tabs.

Security hardening from an independent review on October 31, 2026 (no Critical findings; see Independent security review).

  • tools/create-com-user.py grants only what the tabs need. Earlier copies set Read on every Morpheus permission that wasn't on a short block list: about 45 features on Morpheus 9.0.1, including logs, health, tasks, templates and Terraform state. The role now gets Read on Operations: Dashboard and Infrastructure: Compute, Clusters, Groups, Clouds, default access to groups and clouds, and the plugin's own permissions. Everything else stays at None, and the script stops before creating a user if anything else is above None. If you made a role with the 0.4.2 script, delete it and run the new script, or set the other permissions to None by hand.
  • The client secret and token go only to HPE. The COM API URL, Token URL override and COM UI URL settings must be https:// addresses under greenlake.hpe.com or cloud.hpe.com, with no user name, query or fragment. The workspace ID must be 32 hexadecimal characters, and the proxy URL must be http:// or https:// with a host name. If a setting doesn't qualify, the tabs show which one to fix and the plugin doesn't sign in.
  • Memory stays bounded on large workspaces. Matching keeps only each server's MAC addresses, not its whole inventory. The response cache holds at most 1,000 entries, dropping expired ones first. Responses over 16 MB are refused.
  • Per-user limits on the action routes. Each Morpheus user can make 60 route calls a minute and start 20 actions an hour. An iLO can be reset once every 10 minutes. Job progress is reused for 3 seconds, so polling costs COM one call per interval.
  • Action tokens expire. A host tab's action token works on the day it was issued and the next day. (It isn't tied to the user, because Morpheus doesn't tell a tab who's viewing it. The routes still need the actions permission.)
  • iLO links are made only from a plain IP address or host name reported to COM; anything else is shown as text.
  • Errors: unexpected errors on the host tab and during matching show a reference number instead of the exception text.
  • COM groups are read once an hour in the background, instead of one COM call per group while a page loads.
  • Diagnostics: the unused probeOpen route is gone, and the probe route no longer echoes the caller's name and permission.
  • Build: tools/check-plugin-core.sh works in build/compat instead of a shared /tmp folder. Test data uses example values for NIC serial numbers and storage identifiers.
  • New SECURITY.md: how to report a vulnerability, and which versions get fixes.
  • Docs: the screenshots use example values (host-a, cluster-a, XYZ0000001, 192.0.2.21, workspace my-workspace) instead of blocked-out text, and the same value means the same thing in every image and section. The README shows the plugin in the Morpheus UI: annotated host and cluster pages from a real 9.0.1 appliance. The install guide lists the example values up front.

Faster tabs, smaller COM reads and a quicker index refresh, from measurements on a Morpheus 9.0 appliance (October 31, 2026). Nothing changes in what the tabs show or in the checks on the action routes.

  • Faster tabs. Morpheus asks a tab whether to show itself and then renders it, one right after the other. The host tab now reads the host from Morpheus once per page instead of two or three times (each read took 0.7 to 1.1 s), and the cluster tab lists the cluster's servers once instead of twice (2 to 2.8 s each). What was read is reused for at most 15 seconds. Each tab still checks the viewer's permission and tenant on every page, and the action buttons always read the host fresh.
  • Smaller COM reads. The plugin asks COM for compressed (gzip) answers: a server's inventory drops from about 136 KB to 14 KB. The 16 MB limit applies to the uncompressed size.
  • The cluster summary reads only its own servers from COM, 50 at a time, instead of paging through every server in the workspace. If COM refuses that query, it reads the whole list as before. If a COM list is ever longer than the plugin reads (50 pages), the Morpheus log says so once.
  • Quicker index refresh. Rebuilding the host-to-COM match between the regular 6-hour refreshes reads a server's inventory only when COM has collected a new one since the last build, or after a successful Collect inventory action; the 6-hour refresh still reads every inventory. A host with no match triggers a rebuild at most once an hour, and never when Morpheus has no MAC addresses for it, so a server newly added to COM can take up to an hour to match on a host page that already showed "No COM match". After a failed build the plugin waits longer each time (up to 15 minutes), or as long as COM asks after a rate limit. A build stops if COM can't be reached, and the last good index stays in use.
  • Outages: when sign-in fails or COM can't be reached, tabs show the error at once for the next 45 seconds instead of each waiting for timeouts. Action buttons fail the same way during those 45 seconds. Matching keeps retrying in the background. A broken or cut-off answer fails only that request.
  • Changing the connection settings stops a build that's still using the old ones, and the new connection starts clean (its error state, alert fallback and rebuild limits).
  • Reliability: cached COM answers are plain maps and lists, safe to share between page loads.