feat(pending-updates): short-circuit an unchanged fleet on a DB rowversion watermark (#62)

Owner decision (b), 2026-11-09. PendingUpdatesEvaluator re-flattened every
deployed instance on each cache-TTL expiry even when nothing had changed.

- Migration AddConfigGraphRowVersions: a non-concurrency rowversion column
  + index on the 18 configuration tables the evaluation reads (Templates +
  5 member tables, Instances + 4 override/binding tables, DataConnections,
  DeployedConfigSnapshots, Sites, SharedScripts, ExternalSystemDefinitions,
  ExternalSystemMethods, DatabaseConnectionDefinitions). SQL Server only in
  the model; SQLite test contexts are untouched. Script:
  docs/plans/sql/AddConfigGraphRowVersions.sql. Joins the coordinated
  central-pair migration window (#107 item 4).
- IConfigGraphWatermarkReader / ConfigGraphWatermarkReader: ONE query of
  MAX(RowVersion) + COUNT_BIG(*) per table (count catches deletes, which
  raise no rowversion). rowversion is engine-bumped for any writer, so a
  peer central node's edit moves it — not the per-process watermark MG-3
  removed.
- The secret store (outside the DB) gets its own stamp: SHA-256 over
  ListAsync(includeDeleted: true) metadata.
- Evaluator re-serves the last CLEAN snapshot while watermark + secret
  stamp are unchanged; every PendingUpdatesFullEvaluationEvery'th (default
  10, validated >= 1, 1 = off) is full regardless — the safety net for a
  same-KEK re-wrap or an input table missing from the set. Always full when
  the token is unavailable/unreadable, after an evaluation that could not
  compute a site's artifact hash, on Re-check, and after a deploy completion.
  It only re-serves a real earlier answer, never manufactures one.

Tests: SQL Server (real migrated DB) — token stable without writes, moves
for a peer-connection INSERT/UPDATE and a non-max DELETE, moves for a local
EF write, a detached Update still saves (not a concurrency token), and a
guard that every table the evaluator's repository reads touch is covered.
Evaluator — reuse without flatten (re-served answer still shows the stale
instance), watermark move, 4 secret changes, every-Nth, 5 no-token modes,
unclean evaluation not reused, cache Re-check and deploy completion force
full. Mutation-proven: token compare, clean check, cache force flag, row
count, and an uncovered read each turn tests red.

Claude-Session: https://claude.ai/code/session_01GT67AEHd2Acv1YPwGfXEoV
This commit is contained in:
Joseph Doherty
2026-11-09 22:38:55 -05:00
parent 63ec45e2e3
commit a4806576af
20 changed files with 3963 additions and 9 deletions
@@ -72,8 +72,26 @@ Three Central UI changes, requested together:
(default 60 s, validated > 0), running the evaluator in its own DI scope. `GetAsync(forceRefresh)`.
Consequence: however many circuits poll, the fleet is flattened at most once per TTL, and not at
all when nobody is looking.
- Not in scope (register rows): a template-graph-watermark short-circuit so an unchanged fleet costs
no re-flatten; a `deploy` CLI/`Health` KPI surface for the counts.
- Not in scope (register rows): a `deploy` CLI/`Health` KPI surface for the counts.
- **Unchanged-fleet short-circuit (Gitea #62, register row 124, owner decision (b) 2026-11-09).** Before
the fleet pass the evaluator reads a change token = `IConfigGraphWatermarkReader` (ONE query:
`MAX(RowVersion)` + `COUNT_BIG(*)` per table over the 18 input tables — Templates + their five member
tables, Instances + their four override/binding tables, DataConnections, DeployedConfigSnapshots, Sites,
SharedScripts, ExternalSystemDefinitions, ExternalSystemMethods, DatabaseConnectionDefinitions; migration
`AddConfigGraphRowVersions`, list in `ConfigGraphRowVersions`) **plus** a SHA-256 stamp over the secret
store's `ListAsync(includeDeleted: true)` metadata (name, revision, created/updated, tombstone, KEK id,
content type, description). Equal to the token read before the last *clean* full evaluation → the
previous snapshot is re-served (with a fresh `ComputedAt`). rowversion is database-wide and engine-bumped,
so a peer central node's edit, a bundle import or an `ExecuteUpdate` all move it; the count catches a
delete (which raises no rowversion). It is deliberately not the per-process template-graph watermark
(MG-3). Safety net: every `PendingUpdatesFullEvaluationEvery`th evaluation (default 10 ≈ 10 min at the
60 s TTL; `1` disables the short-circuit) is full regardless — it covers a same-KEK re-wrap (wrap
envelope changes, revision does not) and any input table someone forgets to add (a SQL Server guard
test fails if a repository read used by the evaluator touches an uncovered table). Always full: watermark
unavailable (non-SQL-Server provider) or unreadable, secret listing fails, the previous evaluation could
not compute a site's artifact hash (secret-store outage — re-serving it would pin a transient *Out of
date*), a *Re-check* (`forceRefresh`), and the evaluation after a deploy completion. It never
manufactures an answer — it only re-serves a real earlier one.
### 3.3 Sites page (Admin)
@@ -0,0 +1,84 @@
-- AddConfigGraphRowVersions (2026-11-10, ScadaBridge#62). A rowversion column + index on the 18 configuration
-- tables feeding the pending-updates evaluation, so PendingUpdatesEvaluator can skip the fleet flatten when
-- MAX(RowVersion)+COUNT_BIG(*) over them is unchanged. Additive: SQL Server fills every existing row as the
-- column is added; old binaries ignore the column. Apply after AddAuditLogHashChain in the coordinated
-- central-pair migration window (ScadaBridge#107 item 4).
BEGIN TRANSACTION;
ALTER TABLE [TemplateScripts] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [Templates] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [TemplateNativeAlarmSources] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [TemplateCompositions] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [TemplateAttributes] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [TemplateAlarms] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [Sites] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [SharedScripts] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [Instances] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [InstanceNativeAlarmSourceOverrides] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [InstanceConnectionBindings] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [InstanceAttributeOverrides] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [InstanceAlarmOverrides] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [ExternalSystemMethods] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [ExternalSystemDefinitions] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [DeployedConfigSnapshots] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [DataConnections] ADD [RowVersion] rowversion NOT NULL;
ALTER TABLE [DatabaseConnectionDefinitions] ADD [RowVersion] rowversion NOT NULL;
CREATE INDEX [IX_TemplateScripts_RowVersion] ON [TemplateScripts] ([RowVersion]);
CREATE INDEX [IX_Templates_RowVersion] ON [Templates] ([RowVersion]);
CREATE INDEX [IX_TemplateNativeAlarmSources_RowVersion] ON [TemplateNativeAlarmSources] ([RowVersion]);
CREATE INDEX [IX_TemplateCompositions_RowVersion] ON [TemplateCompositions] ([RowVersion]);
CREATE INDEX [IX_TemplateAttributes_RowVersion] ON [TemplateAttributes] ([RowVersion]);
CREATE INDEX [IX_TemplateAlarms_RowVersion] ON [TemplateAlarms] ([RowVersion]);
CREATE INDEX [IX_Sites_RowVersion] ON [Sites] ([RowVersion]);
CREATE INDEX [IX_SharedScripts_RowVersion] ON [SharedScripts] ([RowVersion]);
CREATE INDEX [IX_Instances_RowVersion] ON [Instances] ([RowVersion]);
CREATE INDEX [IX_InstanceNativeAlarmSourceOverrides_RowVersion] ON [InstanceNativeAlarmSourceOverrides] ([RowVersion]);
CREATE INDEX [IX_InstanceConnectionBindings_RowVersion] ON [InstanceConnectionBindings] ([RowVersion]);
CREATE INDEX [IX_InstanceAttributeOverrides_RowVersion] ON [InstanceAttributeOverrides] ([RowVersion]);
CREATE INDEX [IX_InstanceAlarmOverrides_RowVersion] ON [InstanceAlarmOverrides] ([RowVersion]);
CREATE INDEX [IX_ExternalSystemMethods_RowVersion] ON [ExternalSystemMethods] ([RowVersion]);
CREATE INDEX [IX_ExternalSystemDefinitions_RowVersion] ON [ExternalSystemDefinitions] ([RowVersion]);
CREATE INDEX [IX_DeployedConfigSnapshots_RowVersion] ON [DeployedConfigSnapshots] ([RowVersion]);
CREATE INDEX [IX_DataConnections_RowVersion] ON [DataConnections] ([RowVersion]);
CREATE INDEX [IX_DatabaseConnectionDefinitions_RowVersion] ON [DatabaseConnectionDefinitions] ([RowVersion]);
INSERT INTO [__EFMigrationsHistory] ([MigrationId], [ProductVersion])
VALUES (N'20261110032839_AddConfigGraphRowVersions', N'10.0.13');
COMMIT;
GO
@@ -569,9 +569,23 @@ at once; one run through the peer central clears within the TTL. See
the Sites-page badge/re-check UI and the rail-footer indicator
(Component-CentralUI.md → *Site & Data Connection Management* and *Technology*).
Not in scope for this slice (deferred-work register rows 124–125): a
template-graph-watermark short-circuit so an unchanged fleet costs no re-flatten
per TTL, and a `deploy` CLI / Health-dashboard KPI surface for the counts.
**Unchanged-fleet short-circuit (Gitea #62, register row 124).** Inside a flight the
evaluator first reads a change token — `IConfigGraphWatermarkReader` (one query of
`MAX(RowVersion)` + `COUNT_BIG(*)` over the 18 tables the evaluation reads, the
`rowversion` columns added by migration `AddConfigGraphRowVersions`; the covered set is
`ConfigGraphRowVersions.EntityTypes`) plus a SHA-256 stamp over the secret store's
metadata listing — and re-serves the last clean snapshot when it is unchanged.
`rowversion` is bumped by the engine for any writer, so an edit through the peer central
node is detected (the per-process template-graph watermark would not be — arch-review
2026-09-09 MG-3). Every `ScadaBridge:DeploymentManager:PendingUpdatesFullEvaluationEvery`th
evaluation (default 10, validated ≥ 1; 1 disables the short-circuit) is full regardless,
the safety net for a same-KEK secret re-wrap and for an input table missing from the set.
A *Re-check* (`forceRefresh`), a deploy completion, an unavailable watermark or secret
listing, and an evaluation that could not compute a site's artifact hash all force a full
evaluation. Only a real earlier answer is ever re-served.
Not in scope for this slice (deferred-work register row 125): a `deploy` CLI /
Health-dashboard KPI surface for the counts.
## Instance Lifecycle Commands