Merge #868 fourth review fixes: poison only by SQL error allowlist, memory-counted acks survive restart, Stale-recovered revisions unconfirmed

Claude-Session: https://claude.ai/code/session_01GT67AD99zQQqpLCVm3p7xk
This commit is contained in:
Joseph Doherty
2026-11-11 13:35:42 -05:00
7 changed files with 367 additions and 30 deletions
+7 -1
View File
@@ -687,7 +687,13 @@ deployment green fleet-wide. Now:
whose ConfigDb is unreachable) restores the cached revision — the cache is written before the outcome — and
sends **no ack**, so central times that deployment out even though the node runs it. The restored revision
is marked *unconfirmed*, so **re-deploying it re-applies** and acks the real outcome rather than
short-circuiting a clean `Applied`.
short-circuiting a clean `Applied`. ⚠️ A `Direct` node running from cache that **still cannot reach the
ConfigDb** cannot read the artifact to re-apply, so it acks that same-revision re-dispatch **`Failed`** and
the deployment reads **`PartiallyFailed`**, not `Sealed` — the honest answer; re-deploy once the node reaches
the ConfigDb again.
- The same *unconfirmed* marking applies to a revision adopted by **Stale recovery** (which restores only the
revision marker, not drivers or the address space) and to one whose **bootstrap restore failed**: a dispatch
of it re-applies instead of acking clean over an empty address space.
- **Upgrade note (lmxopcua#868): do not deploy while the central pair is mid rolling-upgrade across this
change.** An old-build fused central node still writes its own row `Applied` before the outcome, which a
new-build coordinator's recovery would count as a clean ack. Finish upgrading both central nodes, then deploy.