← Back to list
01Issue
FeatureShippedSwamp CLIPublic
Assigneesstack72
Relationships
#4811 serve: a team serve restarts one host at a time with no failed requests and no lost runs
Opened by stack72 · 11/16/2026· Shipped 11/17/2026
Follow-up from #4439, which asked: "On a team serve (v3: followers, lease takeover), restart one host at a time behind the load balancer." It waits on team mode, datastore Phase 5.
Context
A single serve now restarts without failed requests or lost runs:
--reuse-porthands the listening socket over (#4566);- a planned restart parks runs at a step boundary and the new process continues them (#4737);
- a crash continues every run that is safe to continue (#4693, #4703).
Overlapping processes on a v3 datastore are unsupported until team mode ships, because both would need the writer lease (design/primitives/serve.md, Rolling restart).
Ask
When team mode ships, restart a team serve one host at a time with no failed requests and no lost runs:
- Restarting a follower: its connections move to the other hosts behind the load balancer. Since a follower owns no runs, nothing is parked.
- Restarting the leader: it parks its runs (#4737) and gives up the writer lease. A follower is promoted and continues the parked runs through the same claim and continuation path, with no window in which writes fail.
- An operator runbook: the order to restart hosts in, what to watch (
/ready, the leader record), and how to roll back.
Tests
- Rolling a three-host team, one host at a time, during a running workflow and a stream of webhook deliveries: no failed delivery, and the workflow finishes with no step run twice.
- Restarting the leader mid-run: the new leader continues the run.
02Bog Flow
Shipped
Click a lifecycle step above to view its details.
03Sludge Pulse
Sign in to post a ripple.