Skip to content

docs: Stop replaces Pause for taking a server offline - #16

Open
MorningLightMountain713 wants to merge 1 commit into
RunOnFlux:masterfrom
MorningLightMountain713:docs/retire-pause-container
Open

docs: Stop replaces Pause for taking a server offline#16
MorningLightMountain713 wants to merge 1 commit into
RunOnFlux:masterfrom
MorningLightMountain713:docs/retire-pause-container

Conversation

@MorningLightMountain713

@MorningLightMountain713 MorningLightMountain713 commented Aug 20, 2026

Copy link
Copy Markdown

Pause and Unpause are being removed from FluxOS. Seven live pages on docs.runonflux.com currently tell customers to use them.

Why pause went

Docker reports a paused container as running. That one fact meant the network could not see the state at all:

  • the load balancer kept routing players to a server whose processes were frozen and which could never answer
  • monitoring charts flatlined with no explanation
  • the reconciler saw a perfectly healthy app and took no action
  • nothing could release one — unpause went too, so a paused container stayed frozen and unrecoverable

Stop does the same job, and does it better

Pause Stop
Releases the world files yes yes
Players can connect meanwhile no — a frozen container answers nothing no
Master stays named on the load balancer yes yes (see dependency below)
Comes back on the same node yes yes
Recoverable afterwards no — no unpause exists yes

Pause never kept the server playable. The one thing it did that Stop did not was hold the load-balancer primary designation, and that gap is closed by the FDM change this depends on.

What changed

Six world-upload guides — terraria, valheim, projectzomboid, minecraftbedrock, windrose, vrising — said pause the container, upload, restart. They now say stop, upload, start, with a note that a g: server can take up to a minute to come back.

Three reference pages — the Control table drops the Pause/Unpause rows and explains why, the management overview no longer lists pause among per-node actions, and the container status list no longer offers "paused" as a state.

Deliberately left alone

I swept all 33 occurrences of "pause" in docs/, not just the Pause control. These are not ours and are unchanged: satisfactory's in-game auto-pause setting, Project Zomboid's PauseEmpty= config, Don't Starve Together's Sim paused log line, three MaxGCPauseMillis Java flags, the billing FAQ's "can I pause my subscription", and the ArcaneOS note about services pausing during setup.

Merge dependencies

Do not publish before both of these are live:

Dependency Why
RunOnFlux/flux#1777 Retires pause. Until it ships, these pages describe a button that still works
RunOnFlux/flux-domain-manager#154merged AND deployed Keeps an operator-stopped master named as the load-balancer primary. Without it deployed, stopping a g: server drops it out of the load balancer and the primary can move to another node's copy of the data on the next start — making the "comes back on the same node" promise above true in the code and false on the fleet

Related: RunOnFlux/fluxos-frontend#140 removes the buttons from the UI, and RunOnFlux/fluxdocs#50 is the matching API-reference change.

Pause and Unpause are being removed from FluxOS. Docker reports a paused
container as running, so the network could not see the state at all: the load
balancer kept routing players to a server whose processes were frozen and which
could never answer, monitoring charts flatlined with no explanation, and the
reconciler saw a perfectly healthy app. Nothing could release one either -
unpause went with it - so a paused container stayed frozen and unrecoverable.

Stop does the same job for every procedure documented here. It releases the
files, it is visible everywhere as stopped, and it can always be started again.
The one thing Pause did that Stop did not - keep the load balancer pointing at
this instance so the server comes back on the same node with the same data - now
holds for Stop as well.

Six world-upload guides said to pause the container before uploading and restart
afterwards: terraria, valheim, projectzomboid, minecraftbedrock, windrose and
vrising. They now say stop and start. The Control reference drops the two rows
and explains why, the management overview no longer lists pause among the
per-node actions, and the container status list no longer offers "paused" as a
state a container can be in.

Left alone deliberately - these are not the FluxOS control: satisfactory's
in-game auto-pause setting, Project Zomboid's PauseEmpty= config, Don't Starve
Together's "Sim paused" log line, the three MaxGCPauseMillis Java flags, the
billing FAQ's pause question, and the ArcaneOS note about services pausing
during setup.

DEPENDS ON the flux-domain-manager change that keeps an operator-stopped master
named as primary. Without it deployed, stopping a g: server drops it out of the
load balancer and the primary can move to another node's copy of the data on the
next start - which would make the "comes back on the same node" promise above
false on the fleet even though it is true in the code. Do not publish ahead of
that deploy.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant