lei.chen/myxce
1
0
Fork 1
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
myxce/docs/superpowers/plans/2026-05-07-persistence-ha-overview.md
Rongsong Shen 59597e105e feat: add --peers to controller for HA peer list and priority election 
- myconf_t gains peers_cfg (xce_cluster_config_t) populated from --peers
- controller/main.c: parse --peers host:port,... via argp; ca_cert_path
  auto-derived from workdir/certs/xce.crt (no --ca-cert flag needed)
- cluster_init: derive my_priority from position in peers list (index 0 =
  primary); fall back to 0 when --peers not given (single-node mode)
2026-05-08 04:40:27 +00:00

2.2 KiB
Raw Blame History

XCE Persistence & HA — Implementation Overview

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Make the XCE controller survive restarts and primary failures with no job/worker/queue state loss.

Architecture: WAL + periodic checkpoint (dual-zarray COW) for single-controller persistence; LSF-style priority election + fan-out TCP replication for HA. libxce smart discovery handles client auto-reconnect after failover.

Tech Stack: C (controller/lib), Cap'n Proto (serialization), Nix flakes + meson + ninja (build), Go (API server discovery bridge)

Phase Map

Each phase is an independently shippable increment. Create the next phase plan only after the current phase passes its gate tests.

Phase Plan file Goal Gate
P0 2026-05-07-p0-single-controller-persistence.md Controller survives restart P0-G smoke tests green
P1 create after P0 ships Standby takes over when primary fails Election + promotion tests green
P2 create after P1 ships Primary replicates to standbys in real time Replication round-trip + catch-up tests green
P2.5 create alongside P2 Workers/schedulers auto-reconnect after failover Discovery + reconnect tests green
P3 create after P2 ships End-to-end failover + hardening + metrics Full failover test + benchmark passing
P4 create after P3 ships Docs, runbook, stale-ref cleanup Runbook review complete

Out of Scope (all phases)

  • Admin CLI for manual failover (xce hostcontrol --failover)
  • WAL compaction beyond checkpoint trimming
  • Multi-DC replication
  • Runtime config changes without restart
  • Snapshot compression
  • config_version hash in heartbeat (deferred to P3 tooling)
  • Degraded-mode restricted-state timeout (repl_degraded_timeout_s)
  • MAX_HA_CANDIDATES > 4 / MAX_STANDBYS > 3

Design Document

src/docs/persistence-and-ha-design.md — all implementation detail lives here. Read the relevant Part before starting each task.