novox/mesh-catalog
3
0
Fork 0
Code Issues Pull Requests 3 Packages Projects Releases Wiki Activity

Manifests publish software ports; the machine side is the assignment's #173

Closed
mesh-admin wants to merge 2 commits from fix/manifests-publish-software-ports into main
pull from: fix/manifests-publish-software-ports
merge into: :main
Conversation 2 Commits 2 Files Changed 5
+13 -13
mesh-admin commented 2026-09-30 10:36:00 +00:00
Contributor
Copy Link
Copy Source

postgres, distribution, gitea, mailu, only-office and portainer published
long-form mappings (5432:5432, 222:22, 7080:80, 9090:9000 …): a machine port
in a definition (ADR 0038, the migration note). Each now publishes the
software's port and names it in listens (mailu's web/web-tls and
only-office's and portainer's web said the machine number); mailu's env
says ${port:80}/${port:443} instead of 7080/7443.

Nothing moves on novox. postgres, distribution and gitea already have their
machine ports as novox settings (6852, 5100, 222); mailu, only-office and
portainer need theirs set BEFORE this rolls out:
settings set mailu {"ports":{"80":7080,"443":7443}} --node novox
settings set only-office {"ports":{"80":9070}} --node novox
settings set portainer {"ports":{"9000":9090,"9443":9443}} --node novox
Those pins are accepted by today's manifests too (GivenPorts answers to the
container port), so setting them first changes nothing either.

Verified with the controller's own publishedOn/portInto/GivenPorts on
novox's settings (+ the three pins): every container's published ports and
mailu's env file render byte-identical before and after — mesh-store
6852:5432, mesh-registry 5100:5000, gitea 222:22, mailu-front 7080:80
7443:443 (+ mail ports), only-office 9070:80, portainer 9090:9000 9443:9443.

Not included: nats publishes 127.0.0.1:8222:8222 — a loopback bind, which a
short form cannot express; its monitor port needs its own change.

postgres, distribution, gitea, mailu, only-office and portainer published long-form mappings (5432:5432, 222:22, 7080:80, 9090:9000 …): a machine port in a definition (ADR 0038, the migration note). Each now publishes the software's port and names it in listens (mailu's web/web-tls and only-office's and portainer's web said the machine number); mailu's env says ${port:80}/${port:443} instead of 7080/7443. Nothing moves on novox. postgres, distribution and gitea already have their machine ports as novox settings (6852, 5100, 222); mailu, only-office and portainer need theirs set BEFORE this rolls out: settings set mailu {"ports":{"80":7080,"443":7443}} --node novox settings set only-office {"ports":{"80":9070}} --node novox settings set portainer {"ports":{"9000":9090,"9443":9443}} --node novox Those pins are accepted by today's manifests too (GivenPorts answers to the container port), so setting them first changes nothing either. Verified with the controller's own publishedOn/portInto/GivenPorts on novox's settings (+ the three pins): every container's published ports and mailu's env file render byte-identical before and after — mesh-store 6852:5432, mesh-registry 5100:5000, gitea 222:22, mailu-front 7080:80 7443:443 (+ mail ports), only-office 9070:80, portainer 9090:9000 9443:9443. Not included: nats publishes 127.0.0.1:8222:8222 — a loopback bind, which a short form cannot express; its monitor port needs its own change.
mesh-admin added 1 commit 2026-09-30 10:36:00 +00:00
Manifests publish software ports; the machine side is the assignment's a49f598a4e 
postgres, distribution, gitea, mailu, only-office and portainer published
long-form mappings (5432:5432, 222:22, 7080:80, 9090:9000 …): a machine port
in a definition (ADR 0038, the migration note). Each now publishes the
software's port and names it in listens (mailu's web/web-tls and
only-office's and portainer's web said the machine number); mailu's env
says ${port:80}/${port:443} instead of 7080/7443.

Nothing moves on novox. postgres, distribution and gitea already have their
machine ports as novox settings (6852, 5100, 222); mailu, only-office and
portainer need theirs set BEFORE this rolls out:
  settings set mailu {"ports":{"80":7080,"443":7443}} --node novox
  settings set only-office {"ports":{"80":9070}} --node novox
  settings set portainer {"ports":{"9000":9090,"9443":9443}} --node novox
Those pins are accepted by today's manifests too (GivenPorts answers to the
container port), so setting them first changes nothing either.

Verified with the controller's own publishedOn/portInto/GivenPorts on
novox's settings (+ the three pins): every container's published ports and
mailu's env file render byte-identical before and after — mesh-store
6852:5432, mesh-registry 5100:5000, gitea 222:22, mailu-front 7080:80
7443:443 (+ mail ports), only-office 9070:80, portainer 9090:9000 9443:9443.

Not included: nats publishes 127.0.0.1:8222:8222 — a loopback bind, which a
short form cannot express; its monitor port needs its own change.
mesh-admin commented 2026-09-30 10:36:26 +00:00
Author
Contributor
Copy Link
Copy Source

Follow-up: gitea is out of this PR. mesh-controller asserts git-over-ssh publishes at 222 by default when no node sets it (TestTheForgesSshPortIsTheMeshsFixedConventionByDefault) — the short form would render 22:22 there and collide with the machine sshd. That is a design decision for the operator, not part of this sweep. Catalogue test on the branch now passes.

Follow-up: gitea is **out** of this PR. mesh-controller asserts git-over-ssh publishes at 222 by default when no node sets it (`TestTheForgesSshPortIsTheMeshsFixedConventionByDefault`) — the short form would render `22:22` there and collide with the machine sshd. That is a design decision for the operator, not part of this sweep. Catalogue test on the branch now passes.
jschoubben added 1 commit 2026-09-30 10:36:26 +00:00
Leave gitea's 222:22 to its own decision 3728b07d20 
A controller test holds git-over-ssh at 222 as the mesh's fixed convention
when no node says otherwise (TestTheForgesSshPortIsTheMeshsFixedConvention
ByDefault): without a setting, the short form would publish 22:22 and
collide with the machine's own sshd. Changing that is a decision, not part
of this sweep.
mesh-admin commented 2026-10-01 09:25:03 +00:00
Author
Contributor
Copy Link
Copy Source

Closing as superseded: stale against main (two conflicts) and its portainer slice landed as #189. Still true for distribution, mailu, only-office and postgres: their manifests carry host:container machine ports; each gets the same one-line change when its node has a window (ports are the mesh's to assign, no pins).

Closing as superseded: stale against main (two conflicts) and its portainer slice landed as #189. Still true for distribution, mailu, only-office and postgres: their manifests carry `host:container` machine ports; each gets the same one-line change when its node has a window (ports are the mesh's to assign, no pins).
mesh-admin closed this pull request 2026-10-01 09:25:04 +00:00

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No Reviewers
Labels
No items
No labels
Milestone
No items
No Milestone
Projects
Clear projects
No projects
Assignees
Clear assignees
jschoubben miel.truyen
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: novox/mesh-catalog#173