Federation: list and remove servers while Open Cloud Mesh is switched off (#1071 follow-up) #1083

Open
opened 2026-11-03 05:32:10 +00:00 by Cordy · 0 comments
Owner

Today. Every admin OCM route is wrapped in ocmGate, which answers 404 while OCM is off (internal/api/ocm.go). An administrator can then neither see the server list nor remove a server.

Why it matters. NIST SP 800-61 orders incident handling as containment, then eradication, then recovery. For a compromised partner server:

  • Switching OCM off is containment; removing the server is eradication.
  • Today the administrator must switch OCM back on (recovery) before removing the server. For that window the server is trusted again.
  • Since v0.6.390 the administrator can stop every share to that server from Sharing › Shares while OCM is off. That closes most of the window, but the server's shares in and its allow-list entry still come back.

Proposal.

  • Let GET and DELETE /api/v1/admin/ocm/servers pass ocmGate; the other routes stay gated.
  • The Servers tab works while OCM is off, with a note saying so.
  • The removal dialog shows its off variant (dgNoteFedNotTold). The removal records told=off per share.

Tests. Both routes answer while OCM is off and the others still answer 404. A removal while off ends the shares and writes share-ended told=off per share. The dialog's off variant passes axe. Run on posix and s3, with two instances (#742).

Strings. Probably one note for the Servers tab, so a mockup is needed first. This also replaces the raw English 404 hint the tab shows today (backlog).

Refs: #1071 decision N3; research-1071-unsettled.md.

**Today.** Every admin OCM route is wrapped in `ocmGate`, which answers 404 while OCM is off (`internal/api/ocm.go`). An administrator can then neither see the server list nor remove a server. **Why it matters.** NIST SP 800-61 orders incident handling as containment, then eradication, then recovery. For a compromised partner server: - Switching OCM off is containment; removing the server is eradication. - Today the administrator must switch OCM back on (recovery) before removing the server. For that window the server is trusted again. - Since v0.6.390 the administrator can stop every share to that server from Sharing › Shares while OCM is off. That closes most of the window, but the server's shares in and its allow-list entry still come back. **Proposal.** - Let `GET` and `DELETE /api/v1/admin/ocm/servers` pass `ocmGate`; the other routes stay gated. - The Servers tab works while OCM is off, with a note saying so. - The removal dialog shows its off variant (`dgNoteFedNotTold`). The removal records `told=off` per share. **Tests.** Both routes answer while OCM is off and the others still answer 404. A removal while off ends the shares and writes `share-ended told=off` per share. The dialog's off variant passes axe. Run on posix and s3, with two instances (#742). **Strings.** Probably one note for the Servers tab, so a mockup is needed first. This also replaces the raw English 404 hint the tab shows today (backlog). Refs: #1071 decision N3; research-1071-unsettled.md.
Sign in to join this conversation.
No labels
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference: Cordy/Cairn#1083
No description provided.