Compare commits

..

3 Commits

Author SHA1 Message Date
Oliver Kieselbach 6d1ff8d7f0 fix: code-review follow-ups (watcher dispose race, revoke-enforcement tests, scan guard)
- ConsoleBypassWatcher: re-check _disposed after watcher.Start() — a Dispose()
  racing the background arm could leave an armed WMI subscription no code path
  could ever stop (leaked until process exit).
- SignalRContext: collapse rapid accessRevoked pushes into one restart via an
  in-flight guard — a second start() on a non-Disconnected connection throws
  and burns the retry budget.
- config/all: negative-cache managed ids a rescue scan proved missing (5 min
  TTL) — one stale grant (offboarded tenant) no longer makes every delegated
  config/all call pay a full table scan.
- DelegatedAdmin/TenantGroup functions: extract testable revoke-enforcement
  cores (RevokeCoreAsync, SetEnabledCoreAsync, UnassignCoreAsync,
  DeleteGroupCoreAsync, ValidateGrantRequest) and cover the security flows:
  404-on-missing-row writes no audit row and fires no disconnect; real
  revoke/disable/unassign/group-delete always cuts live streams; grant
  validation is fail-closed (GUID tenantId, recognized roles only).
- dashboard/page.tsx: drop unused next/link import left by the banner removal.

Tests: backend 3099 passed, web 557 passed, agent ConsoleBypassWatcher suite
passed, tsc clean.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-20 13:13:26 +02:00
Oliver Kieselbach 5c8b273f37 chore: bump skills submodule (system-Timestamp warning in raw-query skills)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-20 12:38:04 +02:00
Oliver Kieselbach da958b5398 fix(backend): business timestamps survive storage migrations (OccurredUtc + RowKey decode)
The 2026-07-18 storage migration reset every row's system Timestamp,
exposing that AuditLogs, OpsEvents and Events displayed, filtered and
retention-swept on it (custom "Timestamp" writes were silent no-ops —
reserved system property). All migrated rows read as 2026-07-18.

- New Helpers/BusinessTimestamp: RowKey decoders (audit !revtick_guid,
  ops bare revtick, event yyyyMMddHHmmssfff_seq), RowKey-range OData
  clause builders (tick-exact, with the mandatory RowKey lt '"' guard
  so no lower bound can ever match legacy bare-GUID audit rows), and
  dual DateTimeOffset/DateTime column reads.
- Writes in all three tables now persist OccurredUtc (custom column,
  copied verbatim by future migrations); no-op ["Timestamp"] removed.
- Read mappers resolve OccurredUtc -> RowKey decode -> system Timestamp;
  GetEarliestSessionEventTimestampAsync uses the same resolution so
  StartedAt/Duration reconciliation can no longer absorb the migration
  date on migrated sessions.
- Date-window filters and retention moved off the system Timestamp onto
  RowKey ranges; ops retention loses its full-table scan and the
  post-migration mass-delete cliff; audit retention runs a temporary
  second pass for legacy GUID rows (TODO: remove after 2027-03-01).
- New GlobalAdmin backfill endpoint POST /api/maintenance/
  backfill-occurred-utc (dry-run by default, batched, idempotent MERGE
  of only OccurredUtc, table=audit|ops) — Events rows are deliberately
  never mutated; their read path decodes the RowKey instead.
- Tests: decoders, clause boundary proofs incl. GUID-safety pins,
  mapper precedence, backfill row decisions (Functions.Tests 3078).
- Docs: docs/backend/business-timestamps.md (system Timestamp is never
  authoritative; migration checklist rule).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-20 11:29:53 +02:00
24 changed files with 1581 additions and 116 deletions
+95
View File
@@ -0,0 +1,95 @@
---
type: Concept
title: Business Timestamps vs. the Azure Tables System Timestamp
description: Why the system Timestamp is never authoritative, how the 2026-07-18 migration exposed that, and the OccurredUtc + RowKey-decode compensation (mappers, RowKey-range date filters, retention, backfill).
resource: /src/Backend/AutopilotMonitor.Functions/Helpers/BusinessTimestamp.cs
tags:
- backend
- table-storage
- audit-log
- ops-events
- events
- migration
timestamp: 2026-07-20T00:00:00+02:00
---
# Problem
Azure Table Storage's `Timestamp` is a **system property**: the server ignores any supplied
value on write and stamps the row's *write* time; every row rewrite (storage-account
migration, MERGE) resets it. Three tables nevertheless treated it as the business event
time — writes like `["Timestamp"] = evt.Timestamp` were silent no-ops, unnoticed because
write time ≈ event time in normal operation.
The 2026-07-18 EU storage migration rewrote every row, so all migrated rows in
**AuditLogs**, **OpsEvents**, and **Events** read as "2026-07-18 ~13:58 UTC": displays were
wrong, date-window filters matched the wrong rows, and retention froze (nothing older than
the migration moment) with a mass-delete cliff one retention period later.
# Rule
**The system Timestamp is never authoritative.** Every table that needs an event time
stores it in the custom column `OccurredUtc` (`BusinessTimestamp.OccurredUtcColumn`).
Custom columns are copied verbatim by `scripts/Migration/Copy-TableStorage.ps1`, so future
migrations preserve chronology. Migration checklist: any table whose reads/filters depend
on a timestamp must carry it as a custom column — audit this before every storage move.
# Recovery sources per table
The original time survives migrations inside the RowKey:
| Table | RowKey scheme | Decodable |
|---|---|---|
| AuditLogs | `!{MaxTicks-ticks:D19}_{guid:N}` (since 2026-05-06) | Yes; legacy bare-GUID rows (pre-2026-05) are lost |
| OpsEvents | `{MaxTicks-ticks:D19}` | Yes, all rows |
| Events | `{ts:yyyyMMddHHmmssfff}_{seq:D10}` (sanitized agent time) | Yes, all rows (ms precision) |
# Mechanism
`BusinessTimestamp` (Functions/Helpers) is the single source of truth: RowKey decoders,
the OData clause builders, and `GetUtcDateTime` (datetime columns materialize as
`DateTimeOffset` *or* `DateTime` depending on payload shape — always read via this helper).
1. **Read mappers** resolve `OccurredUtc` → RowKey decode → system Timestamp (last
resort; only audit legacy GUID rows end there). Resolvers:
`TableStorageService.ResolveAuditTimestamp`, `TableOpsEventRepository.ResolveTimestamp`,
`TableStorageService.ResolveEventTimestamp` (also used by
`GetEarliestSessionEventTimestampAsync`, which feeds StartedAt/Duration reconciliation —
a system-Timestamp read there would rewrite migrated sessions' derived fields).
2. **Date-window filters and retention use RowKey ranges**, not property comparisons:
OData excludes rows missing a compared property (an `OccurredUtc` filter would drop
every not-yet-backfilled row), and RowKey ranges are index-backed. Boundary semantics
are tick-exact equivalents of the old `Timestamp ge/le` clauses.
**Trap:** audit legacy GUID RowKeys (hex first char) sort *after* all `!`-rows, so
every lower bound `RowKey ge '!…'` must be paired with `RowKey lt '"'`
(`AuditTimeEncodedUpperBound`) — without the guard, a retention sweep would delete all
legacy rows. Pinned by `BusinessTimestampFilterClauseTests`.
3. **Writes** in all three tables now set `OccurredUtc`; the no-op `["Timestamp"]`
entries were removed.
4. **Backfill** (`POST /api/maintenance/backfill-occurred-utc`, GlobalAdminOnly, dry-run
by default, `table=audit|ops`, batched via `maxRows`/`continuation`): MERGE that adds
only `OccurredUtc` decoded from the RowKey — deterministic and idempotent. **Events
rows are deliberately never backfilled** (read-path decode is sufficient; no bulk
mutation of enrollment evidence). Note: a MERGE resets the row's system Timestamp —
irrelevant once nothing depends on it.
5. **Audit retention** runs two passes until ~2027-01: RowKey-range for `!`-rows plus a
`RowKey ge '0' and Timestamp lt …` pass that ages legacy GUID rows out via their frozen
migration-date system Timestamp (remove after 2027-03-01).
# Consequences
* Audit legacy GUID rows have "unknown date" semantics: excluded from date-filtered
views, shown with the (wrong but honest) 2026-07-18 fallback in unfiltered views.
* Raw endpoints (`/api/raw/*`, `query_table`) intentionally keep showing the raw system
Timestamp — they are storage-state inspectors, not business views.
* Event timestamps for pre-cutover rows resolve to the RowKey's sanitized agent time
(ms precision); `ReceivedAt` (server ingest time, rows since 2026-03-19) and
`OriginalTimestamp` (pre-clamp agent time, clamped rows only) remain separate fields.
# Citations
* `src/Backend/AutopilotMonitor.Functions/Helpers/BusinessTimestamp.cs` — decoders + clause builders
* `src/Backend/AutopilotMonitor.Functions/Services/TableStorageService.Maintenance.cs` — audit write/mapper/filters/retention
* `src/Backend/AutopilotMonitor.Functions/DataAccess/TableStorage/TableOpsEventRepository.cs` — ops write/mapper/filters/retention
* `src/Backend/AutopilotMonitor.Functions/Services/TableStorageService.Sessions.cs` — event writes, `ResolveEventTimestamp`, earliest-event read
* `src/Backend/AutopilotMonitor.Functions/Services/OccurredUtcBackfillService.cs` + `Functions/Admin/BackfillOccurredUtcFunction.cs` — backfill
+4
View File
@@ -23,6 +23,10 @@ https://docs.autopilotmonitor.com (separate repository).
* [Gather Rule Phase Scoping & Emit-on-Change](rules/gather-rule-phase-scoping.md) - Restricting gather rules to enrollment phases (activePhases / activeFromPhase sticky latch) and emitMode "on_change" result-dedup — the anti-spam pair for interval rules.
# Backend
* [Business Timestamps](backend/business-timestamps.md) - Why the Azure Tables system Timestamp is never authoritative (migrations reset it), and the OccurredUtc + RowKey-decode compensation for AuditLogs/OpsEvents/Events after the 2026-07-18 migration.
# Security & Identity
* [MCP OAuth Flow](mcp-oauth-flow.md) - Who authenticates where when connecting an AI client to the MCP server; two identities, three parties plus Entra ID.
+1
View File
@@ -2,6 +2,7 @@
## 2026-07-20
* **Creation**: Added `backend/business-timestamps.md` (new `backend/` section) — the 2026-07-18 migration reset every row's system Timestamp, exposing that AuditLogs/OpsEvents/Events had (unknowingly) displayed and filtered on it all along; documents the `OccurredUtc` column, the RowKey decoders and RowKey-range date/retention clauses in `BusinessTimestamp`, the GUID-guard trap, the audit/ops backfill endpoint, and the "system Timestamp is never authoritative" migration rule.
* **Creation**: Added `rules/gather-rule-phase-scoping.md` (new `rules/` section) — gather rules gain `activePhases` (run only during listed phases), `activeFromPhase` (sticky latch from a phase onwards, never via Failed) and `emitMode: "on_change"` (canonical-hash result dedup with `suppressedPolls` observability), ConfigVersion 34. Includes the prerequisite backend fix: the custom-rule toggle PUT no longer wipes the stored rule definition (partial-merge in `GatherRuleService.UpdateRuleAsync`).
## 2026-07-18
@@ -136,6 +136,25 @@ namespace AutopilotMonitor.Agent.V2.Core.Monitoring.Telemetry.Security
}
watcher.Start();
// Re-check after arming: Dispose() may have run between the lock release above and
// Start() — it then stopped a not-yet-armed watcher and this thread armed it right
// back, leaving a WMI subscription no code path could ever reach again. If disposal
// won that race, tear the watcher down here.
bool disposedDuringArm;
lock (_lock)
{
disposedDuringArm = _disposed;
if (disposedDuringArm) _startWatcher = null;
}
if (disposedDuringArm)
{
try { watcher.EventArrived -= OnProcessStartTrace; } catch { }
try { watcher.Stop(); } catch { }
try { watcher.Dispose(); } catch { }
return;
}
_logger.Info($"[ConsoleBypassWatcher] watching Win32_ProcessStartTrace for " +
$"{string.Join(", ", WatchedProcessNames)} (interactive-session + bare command line)");
}
@@ -0,0 +1,83 @@
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Services;
namespace AutopilotMonitor.Functions.Tests;
/// <summary>
/// RowKey → business-time decoders. These are the read-path compensation for rows whose
/// system Timestamp was reset by the 2026-07-18 storage migration: the RowKey is the only
/// time carrier that survives a row rewrite, so the decode must round-trip exactly and
/// must reject every foreign format (falling back instead of mis-dating rows).
/// </summary>
public class BusinessTimestampDecodeTests
{
private static readonly DateTime Ts = new DateTime(2026, 6, 15, 10, 30, 45, 123, DateTimeKind.Utc).AddTicks(4567);
[Fact]
public void Audit_rowkey_roundtrips_tick_exact()
{
var rowKey = TableStorageService.BuildAuditLogRowKey(Ts, Guid.NewGuid());
Assert.True(BusinessTimestamp.TryDecodeAuditRowKey(rowKey, out var decoded));
Assert.Equal(Ts, decoded);
Assert.Equal(DateTimeKind.Utc, decoded.Kind);
}
[Theory]
[InlineData(null)]
[InlineData("")]
[InlineData("!")]
[InlineData("!123")] // too short
[InlineData("!123456789012345678x_guid")] // non-digit in revtick
[InlineData("d6f6dd90f41c4b16b290aac042e5b466")] // legacy bare GUID (:N)
[InlineData("0000000000000000000")] // ops-style bare revtick is not an audit key
public void Audit_decoder_rejects_foreign_formats(string? rowKey)
{
Assert.False(BusinessTimestamp.TryDecodeAuditRowKey(rowKey, out _));
}
[Fact]
public void Ops_rowkey_roundtrips_tick_exact()
{
var rowKey = $"{DateTime.MaxValue.Ticks - Ts.Ticks:D19}";
Assert.True(BusinessTimestamp.TryDecodeOpsRowKey(rowKey, out var decoded));
Assert.Equal(Ts, decoded);
Assert.Equal(DateTimeKind.Utc, decoded.Kind);
}
[Theory]
[InlineData(null)]
[InlineData("")]
[InlineData("123456789012345678")] // 18 digits
[InlineData("12345678901234567890")] // 20 digits
[InlineData("123456789012345678x")] // non-digit
[InlineData("!1234567890123456789")] // audit-prefixed
public void Ops_decoder_rejects_foreign_formats(string? rowKey)
{
Assert.False(BusinessTimestamp.TryDecodeOpsRowKey(rowKey, out _));
}
[Fact]
public void Event_rowkey_prefix_decodes_to_millisecond_truncated_agent_time()
{
var rowKey = $"{Ts:yyyyMMddHHmmssfff}_{42L:D10}";
Assert.True(BusinessTimestamp.TryDecodeEventRowKeyPrefix(rowKey, out var decoded));
// The RowKey format truncates sub-millisecond ticks at write time.
var expected = new DateTime(2026, 6, 15, 10, 30, 45, 123, DateTimeKind.Utc);
Assert.Equal(expected, decoded);
Assert.Equal(DateTimeKind.Utc, decoded.Kind);
}
[Theory]
[InlineData(null)]
[InlineData("")]
[InlineData("20260615103045123")] // no separator/sequence
[InlineData("2026061510304512_0000000042")] // 16-digit prefix
[InlineData("202606151030451234_0000000042")] // 18-digit prefix
[InlineData("2026061510304512x_0000000042")] // non-digit in prefix
[InlineData("20261315103045123_0000000042")] // month 13
[InlineData("d6f6dd90f41c4b16b290aac042e5b466")] // GUID
public void Event_decoder_rejects_foreign_formats(string? rowKey)
{
Assert.False(BusinessTimestamp.TryDecodeEventRowKeyPrefix(rowKey, out _));
}
}
@@ -0,0 +1,181 @@
using System.Text.RegularExpressions;
using AutopilotMonitor.Functions.DataAccess.TableStorage;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Services;
namespace AutopilotMonitor.Functions.Tests;
/// <summary>
/// RowKey-range date-window / retention clauses. Date filtering moved off the system
/// Timestamp (reset by storage migrations) onto the reverse-tick RowKey; these tests pin
/// (a) tick-exact inclusive/exclusive boundary semantics identical to the old
/// `Timestamp ge/le` clauses, and (b) the GUID-safety invariant: legacy bare-GUID audit
/// rows sort AFTER all '!'-rows, so any lower RowKey bound without the `RowKey lt '"'`
/// guard would match — and a retention sweep would DELETE — every legacy row.
/// </summary>
public class BusinessTimestampFilterClauseTests
{
private static readonly DateTime T = new DateTime(2026, 6, 15, 12, 0, 0, DateTimeKind.Utc).AddTicks(789);
private static readonly string Tenant = "11111111-1111-1111-1111-111111111111";
private static string AuditRk(DateTime ts) => TableStorageService.BuildAuditLogRowKey(ts, Guid.NewGuid());
private static string OpsRk(DateTime ts) => $"{DateTime.MaxValue.Ticks - ts.Ticks:D19}";
/// <summary>Evaluates a clause of the form "RowKey op 'literal' [and RowKey op 'literal']…" against a RowKey.</summary>
private static bool Matches(string clause, string rowKey)
{
var parts = Regex.Matches(clause, @"RowKey (ge|gt|le|lt) '([^']*)'");
Assert.True(parts.Count > 0, $"no RowKey comparisons in clause: {clause}");
foreach (Match m in parts)
{
var cmp = string.CompareOrdinal(rowKey, m.Groups[2].Value);
var ok = m.Groups[1].Value switch
{
"ge" => cmp >= 0,
"gt" => cmp > 0,
"le" => cmp <= 0,
"lt" => cmp < 0,
_ => throw new InvalidOperationException(),
};
if (!ok) return false;
}
return true;
}
// ===== Audit boundaries =====
[Fact]
public void Audit_dateFrom_is_inclusive_to_the_tick()
{
var clause = BusinessTimestamp.AuditDateFromClause(T);
Assert.True(Matches(clause, AuditRk(T))); // ts == from → included
Assert.True(Matches(clause, AuditRk(T.AddTicks(1)))); // newer → included
Assert.False(Matches(clause, AuditRk(T.AddTicks(-1)))); // 1 tick older → excluded
}
[Fact]
public void Audit_dateTo_is_inclusive_to_the_tick()
{
var clause = BusinessTimestamp.AuditDateToClause(T);
Assert.True(Matches(clause, AuditRk(T))); // ts == to → included
Assert.True(Matches(clause, AuditRk(T.AddTicks(-1)))); // older → included
Assert.False(Matches(clause, AuditRk(T.AddTicks(1)))); // 1 tick newer → excluded
}
[Fact]
public void Audit_retention_is_strictly_older_than_cutoff()
{
var clause = BusinessTimestamp.AuditRetentionClause(T);
Assert.False(Matches(clause, AuditRk(T))); // ts == cutoff → kept
Assert.True(Matches(clause, AuditRk(T.AddTicks(-1)))); // 1 tick older → deleted
Assert.False(Matches(clause, AuditRk(T.AddTicks(1)))); // newer → kept
}
// ===== Audit GUID safety (the single most dangerous invariant of this change) =====
[Fact]
public void Legacy_guid_rowkeys_sort_after_the_time_encoded_guard()
{
for (var i = 0; i < 20; i++)
{
var guidRk = Guid.NewGuid().ToString("N");
Assert.True(string.CompareOrdinal(guidRk, BusinessTimestamp.AuditTimeEncodedUpperBound) > 0);
}
}
[Fact]
public void No_audit_date_or_retention_clause_matches_legacy_guid_rowkeys()
{
var clauses = new[]
{
BusinessTimestamp.AuditDateFromClause(T),
BusinessTimestamp.AuditDateToClause(T),
BusinessTimestamp.AuditDateToClause(DateTime.MaxValue.AddTicks(-1)), // widest possible window
BusinessTimestamp.AuditRetentionClause(T),
BusinessTimestamp.AuditRetentionClause(DateTime.MaxValue.AddTicks(-1)), // widest possible sweep
};
foreach (var clause in clauses)
{
for (var i = 0; i < 20; i++)
{
Assert.False(Matches(clause, Guid.NewGuid().ToString("N")),
$"clause matched a legacy GUID RowKey: {clause}");
}
}
}
[Fact]
public void Audit_lower_bounded_clauses_carry_the_time_encoded_guard()
{
Assert.Contains($"RowKey lt '{BusinessTimestamp.AuditTimeEncodedUpperBound}'",
BusinessTimestamp.AuditDateToClause(T));
Assert.Contains($"RowKey lt '{BusinessTimestamp.AuditTimeEncodedUpperBound}'",
BusinessTimestamp.AuditRetentionClause(T));
}
// ===== Ops boundaries =====
[Fact]
public void Ops_dateFrom_is_inclusive_to_the_tick()
{
var clause = BusinessTimestamp.OpsDateFromClause(T);
Assert.True(Matches(clause, OpsRk(T)));
Assert.True(Matches(clause, OpsRk(T.AddTicks(1))));
Assert.False(Matches(clause, OpsRk(T.AddTicks(-1))));
}
[Fact]
public void Ops_dateTo_is_inclusive_to_the_tick()
{
var clause = BusinessTimestamp.OpsDateToClause(T);
Assert.True(Matches(clause, OpsRk(T)));
Assert.True(Matches(clause, OpsRk(T.AddTicks(-1))));
Assert.False(Matches(clause, OpsRk(T.AddTicks(1))));
}
[Fact]
public void Ops_retention_is_strictly_older_than_cutoff()
{
var clause = BusinessTimestamp.OpsRetentionClause(T);
Assert.False(Matches(clause, OpsRk(T)));
Assert.True(Matches(clause, OpsRk(T.AddTicks(-1))));
Assert.False(Matches(clause, OpsRk(T.AddTicks(1))));
}
// ===== Composition into the filter builders =====
[Fact]
public void Audit_filter_builders_emit_rowkey_windows_not_timestamp_clauses()
{
var from = new DateTime(2026, 6, 1, 0, 0, 0, DateTimeKind.Utc);
var to = new DateTime(2026, 7, 1, 0, 0, 0, DateTimeKind.Utc);
var plain = TableStorageService.BuildAuditLogFilter(Tenant, from, to);
var fanOut = TableStorageService.BuildAuditLogFilterWithRowKeyBound(
Tenant, from, to, lastRowKey: "!0123_x", excludeDeletions: false);
foreach (var f in new[] { plain!, fanOut })
{
Assert.DoesNotContain("Timestamp ge", f);
Assert.DoesNotContain("Timestamp le", f);
Assert.Contains(BusinessTimestamp.AuditDateFromClause(from), f);
Assert.Contains(BusinessTimestamp.AuditDateToClause(to), f);
}
Assert.Contains("RowKey gt '!0123_x'", fanOut); // fan-out cursor composes with the window
}
[Fact]
public void Ops_filter_builder_emits_rowkey_window_not_timestamp_clauses()
{
var from = new DateTime(2026, 6, 1, 0, 0, 0, DateTimeKind.Utc);
var to = new DateTime(2026, 7, 1, 0, 0, 0, DateTimeKind.Utc);
var f = TableOpsEventRepository.BuildFilter("Security", from, to);
Assert.NotNull(f);
Assert.DoesNotContain("Timestamp ge", f);
Assert.DoesNotContain("Timestamp le", f);
Assert.Contains(BusinessTimestamp.OpsDateFromClause(from), f);
Assert.Contains(BusinessTimestamp.OpsDateToClause(to), f);
}
}
@@ -0,0 +1,118 @@
using Azure.Data.Tables;
using AutopilotMonitor.Functions.DataAccess.TableStorage;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Services;
namespace AutopilotMonitor.Functions.Tests;
/// <summary>
/// Read-mapper resolution order for the three affected tables:
/// OccurredUtc column → RowKey decode → system Timestamp. The last hop is the
/// migration-corrupted value (a row rewrite resets it), so it must only ever win when
/// the row carries no better source (audit legacy bare-GUID rows).
/// </summary>
public class BusinessTimestampPrecedenceTests
{
private static readonly DateTime Occurred = new(2026, 5, 10, 8, 0, 0, DateTimeKind.Utc);
private static readonly DateTime RowKeyTime = new(2026, 6, 20, 9, 30, 0, DateTimeKind.Utc);
private static readonly DateTimeOffset SystemTs = new(2026, 7, 18, 13, 58, 0, TimeSpan.Zero); // migration moment
private static TableEntity WithSystemTimestamp(TableEntity e)
{
e["Timestamp"] = SystemTs;
return e;
}
// ===== Audit =====
[Fact]
public void Audit_occurredutc_wins_over_rowkey_and_system()
{
var e = WithSystemTimestamp(new TableEntity("t", TableStorageService.BuildAuditLogRowKey(RowKeyTime, Guid.NewGuid()))
{
[BusinessTimestamp.OccurredUtcColumn] = new DateTimeOffset(Occurred),
});
Assert.Equal(Occurred, TableStorageService.ResolveAuditTimestamp(e));
}
[Fact]
public void Audit_rowkey_decode_wins_over_system_when_column_missing()
{
var e = WithSystemTimestamp(new TableEntity("t", TableStorageService.BuildAuditLogRowKey(RowKeyTime, Guid.NewGuid())));
Assert.Equal(RowKeyTime, TableStorageService.ResolveAuditTimestamp(e));
}
[Fact]
public void Audit_legacy_guid_row_falls_back_to_system_timestamp()
{
var e = WithSystemTimestamp(new TableEntity("t", Guid.NewGuid().ToString("N")));
Assert.Equal(SystemTs.UtcDateTime, TableStorageService.ResolveAuditTimestamp(e));
}
// ===== Ops =====
[Fact]
public void Ops_occurredutc_wins_over_rowkey_and_system()
{
var e = WithSystemTimestamp(new TableEntity("Security", $"{DateTime.MaxValue.Ticks - RowKeyTime.Ticks:D19}")
{
[BusinessTimestamp.OccurredUtcColumn] = new DateTimeOffset(Occurred),
});
Assert.Equal(Occurred, TableOpsEventRepository.ResolveTimestamp(e));
}
[Fact]
public void Ops_rowkey_decode_wins_over_system_when_column_missing()
{
var e = WithSystemTimestamp(new TableEntity("Security", $"{DateTime.MaxValue.Ticks - RowKeyTime.Ticks:D19}"));
Assert.Equal(RowKeyTime, TableOpsEventRepository.ResolveTimestamp(e));
}
[Fact]
public void Ops_undecodable_rowkey_falls_back_to_system_timestamp()
{
var e = WithSystemTimestamp(new TableEntity("Security", "not-a-revtick"));
Assert.Equal(SystemTs.UtcDateTime, TableOpsEventRepository.ResolveTimestamp(e));
}
// ===== Events =====
private static string EventRowKey(DateTime ts, long seq = 42) => $"{ts:yyyyMMddHHmmssfff}_{seq:D10}";
[Fact]
public void Event_occurredutc_wins_over_rowkey_and_system()
{
var e = WithSystemTimestamp(new TableEntity("tenant_session", EventRowKey(RowKeyTime))
{
[BusinessTimestamp.OccurredUtcColumn] = new DateTimeOffset(Occurred),
});
Assert.Equal(Occurred, TableStorageService.ResolveEventTimestamp(e));
}
[Fact]
public void Event_rowkey_prefix_wins_over_system_when_column_missing()
{
// The pre-cutover case: no OccurredUtc column, system Timestamp = migration moment.
var e = WithSystemTimestamp(new TableEntity("tenant_session", EventRowKey(RowKeyTime)));
Assert.Equal(RowKeyTime, TableStorageService.ResolveEventTimestamp(e));
}
[Fact]
public void Event_undecodable_rowkey_falls_back_to_system_timestamp()
{
var e = WithSystemTimestamp(new TableEntity("tenant_session", "weird-rowkey"));
Assert.Equal(SystemTs.UtcDateTime, TableStorageService.ResolveEventTimestamp(e));
}
[Fact]
public void Event_occurredutc_read_tolerates_datetime_materialization()
{
// SDK reads surface Edm.DateTime as DateTimeOffset or DateTime depending on payload
// shape — the resolver must accept both (dual-reader lesson).
var e = new TableEntity("tenant_session", EventRowKey(RowKeyTime))
{
[BusinessTimestamp.OccurredUtcColumn] = Occurred, // plain DateTime, not DateTimeOffset
};
Assert.Equal(Occurred, TableStorageService.ResolveEventTimestamp(e));
}
}
@@ -129,4 +129,30 @@ public class GetAllTenantConfigurationsFunctionTests
[Fact]
public void DelegatedBareArrayView_Empty_Empty()
=> Assert.Empty(GetAllTenantConfigurationsFunction.DelegatedBareArrayView(new TenantConfiguration[0]));
[Fact]
public void MissingManagedIds_ReturnsOnlyIdsWithoutConfig_CaseInsensitive()
{
var configs = new List<TenantConfiguration>
{
new() { TenantId = TenantB.ToUpperInvariant() }, // case variant still counts as present
};
var missing = GetAllTenantConfigurationsFunction.MissingManagedIds(new[] { TenantB, TenantC }, configs);
var only = Assert.Single(missing);
Assert.Equal(TenantC, only);
}
[Fact]
public void MissingManagedIds_AllPresent_Empty()
=> Assert.Empty(GetAllTenantConfigurationsFunction.MissingManagedIds(
new[] { TenantB },
new List<TenantConfiguration> { new() { TenantId = TenantB } }));
[Fact]
public void MissingConfigCacheKey_LowercasesId_SoCaseVariantsShareOneVerdict()
=> Assert.Equal(
GetAllTenantConfigurationsFunction.MissingConfigCacheKey(TenantB),
GetAllTenantConfigurationsFunction.MissingConfigCacheKey(TenantB.ToUpperInvariant()));
}
@@ -0,0 +1,61 @@
using Azure.Data.Tables;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Services;
namespace AutopilotMonitor.Functions.Tests;
/// <summary>
/// Per-row backfill decision (pure function): rows that already carry OccurredUtc are
/// skipped (idempotent re-runs), decodable RowKeys yield the deterministic decoded value,
/// anything else is reported undecodable — never guessed.
/// </summary>
public class OccurredUtcBackfillServiceTests
{
private static readonly DateTime Ts = new(2026, 6, 20, 9, 30, 0, DateTimeKind.Utc);
[Fact]
public void Audit_row_with_existing_column_is_skipped()
{
var e = new TableEntity("t", TableStorageService.BuildAuditLogRowKey(Ts, Guid.NewGuid()))
{
[BusinessTimestamp.OccurredUtcColumn] = new DateTimeOffset(Ts),
};
Assert.Equal(OccurredUtcBackfillService.RowDecision.AlreadySet,
OccurredUtcBackfillService.DecideRow(OccurredUtcBackfillService.TableAudit, e, out _));
}
[Fact]
public void Audit_time_encoded_row_yields_decoded_write()
{
var e = new TableEntity("t", TableStorageService.BuildAuditLogRowKey(Ts, Guid.NewGuid()));
var decision = OccurredUtcBackfillService.DecideRow(OccurredUtcBackfillService.TableAudit, e, out var occurred);
Assert.Equal(OccurredUtcBackfillService.RowDecision.Write, decision);
Assert.Equal(Ts, occurred);
}
[Fact]
public void Audit_legacy_guid_row_is_undecodable_not_guessed()
{
var e = new TableEntity("t", Guid.NewGuid().ToString("N"));
Assert.Equal(OccurredUtcBackfillService.RowDecision.Undecodable,
OccurredUtcBackfillService.DecideRow(OccurredUtcBackfillService.TableAudit, e, out _));
}
[Fact]
public void Ops_row_yields_decoded_write_and_existing_column_is_skipped()
{
var rowKey = $"{DateTime.MaxValue.Ticks - Ts.Ticks:D19}";
var bare = new TableEntity("Security", rowKey);
var decision = OccurredUtcBackfillService.DecideRow(OccurredUtcBackfillService.TableOps, bare, out var occurred);
Assert.Equal(OccurredUtcBackfillService.RowDecision.Write, decision);
Assert.Equal(Ts, occurred);
var backfilled = new TableEntity("Security", rowKey)
{
[BusinessTimestamp.OccurredUtcColumn] = new DateTimeOffset(Ts),
};
Assert.Equal(OccurredUtcBackfillService.RowDecision.AlreadySet,
OccurredUtcBackfillService.DecideRow(OccurredUtcBackfillService.TableOps, backfilled, out _));
}
}
@@ -0,0 +1,269 @@
using AutopilotMonitor.Functions.Functions.Admin;
using AutopilotMonitor.Functions.Services;
using AutopilotMonitor.Shared;
using AutopilotMonitor.Shared.DataAccess;
using Microsoft.Extensions.Caching.Memory;
using Microsoft.Extensions.Logging.Abstractions;
using Moq;
using Xunit;
namespace AutopilotMonitor.Functions.Tests;
/// <summary>
/// Covers the SignalR revoke-enforcement cores of <see cref="DelegatedAdminManagementFunction"/> and
/// <see cref="TenantGroupManagementFunction"/> (extracted seams — the HTTP entry points are intentionally
/// not exercised, same rationale as <see cref="GetAllBlockedDevicesFunctionTests"/>). Locks in three
/// security invariants: (1) a miss (no assignment row) neither audits nor disconnects — a typo must not
/// write false customer-visible "access removed" rows while the real grant stays live; (2) a real
/// revoke/disable/unassign/group-delete always cuts the affected UPN's live streams (group authorization
/// is join-time only); (3) grant validation is fail-closed (GUID tenantId, recognized roles only).
/// </summary>
public class RevokeEnforcementTests
{
private const string Upn = "MSP-Admin@Partner.Example";
private const string UpnLower = "msp-admin@partner.example";
private const string TenantA = "11111111-1111-1111-1111-111111111111";
private const string TenantB = "22222222-2222-2222-2222-222222222222";
private const string GroupId = "tpl-aaaaaaaa";
private static DelegatedAdminService BuildService(Mock<IAdminRepository> repo) =>
new(
repo.Object,
new StubTenantEntitlementService(AutopilotMonitor.Functions.Security.TenantEdition.Enterprise),
new MemoryCache(new MemoryCacheOptions()),
NullLogger<DelegatedAdminService>.Instance);
private static Mock<IMaintenanceRepository> BuildAuditRepo()
{
var audit = new Mock<IMaintenanceRepository>();
audit.Setup(a => a.LogAuditEntryAsync(
It.IsAny<string>(), It.IsAny<string>(), It.IsAny<string>(),
It.IsAny<string>(), It.IsAny<string>(), It.IsAny<Dictionary<string, string>?>()))
.ReturnsAsync(true);
return audit;
}
private static void VerifyNoAudit(Mock<IMaintenanceRepository> audit) =>
audit.Verify(a => a.LogAuditEntryAsync(
It.IsAny<string>(), It.IsAny<string>(), It.IsAny<string>(),
It.IsAny<string>(), It.IsAny<string>(), It.IsAny<Dictionary<string, string>?>()), Times.Never);
private static (DelegatedAdminManagementFunction Fn, Mock<IAdminRepository> Repo,
Mock<IMaintenanceRepository> Audit, FakeSignalRNotificationService SignalR) BuildAdminFn()
{
var repo = new Mock<IAdminRepository>();
var audit = BuildAuditRepo();
var signalR = new FakeSignalRNotificationService();
var fn = new DelegatedAdminManagementFunction(
NullLogger<DelegatedAdminManagementFunction>.Instance, BuildService(repo), audit.Object, signalR);
return (fn, repo, audit, signalR);
}
private static (TenantGroupManagementFunction Fn, Mock<IAdminRepository> Repo,
Mock<IMaintenanceRepository> Audit, FakeSignalRNotificationService SignalR) BuildGroupFn()
{
var repo = new Mock<IAdminRepository>();
// Loose-mock defaults return a null Task — every repo read the flows touch needs a concrete setup.
repo.Setup(r => r.GetTenantGroupAsync(It.IsAny<string>())).ReturnsAsync((TenantGroup?)null);
repo.Setup(r => r.GetGroupAssignmentsForUpnAsync(It.IsAny<string>()))
.ReturnsAsync(new List<TenantGroupAssignment>());
repo.Setup(r => r.GetGroupAssigneesAsync(It.IsAny<string>()))
.ReturnsAsync(new List<TenantGroupAssignment>());
var audit = BuildAuditRepo();
var signalR = new FakeSignalRNotificationService();
var fn = new TenantGroupManagementFunction(
NullLogger<TenantGroupManagementFunction>.Instance, BuildService(repo), audit.Object, signalR);
return (fn, repo, audit, signalR);
}
// --- Delegated grant validation (fail-closed) ---
[Fact]
public void ValidateGrant_NullBody_Rejected()
=> Assert.Equal("upn and tenantId are required",
DelegatedAdminManagementFunction.ValidateGrantRequest(null, out _));
[Theory]
[InlineData("", TenantA)]
[InlineData(" ", TenantA)]
[InlineData(Upn, "")]
public void ValidateGrant_MissingFields_Rejected(string upn, string tenantId)
=> Assert.Equal("upn and tenantId are required",
DelegatedAdminManagementFunction.ValidateGrantRequest(
new GrantDelegatedAdminRequest { Upn = upn, TenantId = tenantId }, out _));
[Theory]
[InlineData("contoso.com")]
[InlineData("not-a-guid")]
[InlineData("11111111-1111-1111-1111")]
public void ValidateGrant_NonGuidTenantId_Rejected(string tenantId)
=> Assert.Equal("a valid tenantId (GUID) is required",
DelegatedAdminManagementFunction.ValidateGrantRequest(
new GrantDelegatedAdminRequest { Upn = Upn, TenantId = tenantId }, out _));
[Fact]
public void ValidateGrant_UnknownRole_Rejected()
{
var error = DelegatedAdminManagementFunction.ValidateGrantRequest(
new GrantDelegatedAdminRequest { Upn = Upn, TenantId = TenantA, Role = "GlobalAdmin" }, out _);
Assert.NotNull(error);
Assert.Contains(Constants.DelegatedRoles.DelegatedReader, error);
}
[Fact]
public void ValidateGrant_EmptyRole_DefaultsToLeastPrivilegedReader()
{
Assert.Null(DelegatedAdminManagementFunction.ValidateGrantRequest(
new GrantDelegatedAdminRequest { Upn = Upn, TenantId = TenantA }, out var role));
Assert.Equal(Constants.DelegatedRoles.DelegatedReader, role);
}
[Fact]
public void ValidateGrant_ExplicitDelegatedAdmin_Accepted()
{
Assert.Null(DelegatedAdminManagementFunction.ValidateGrantRequest(
new GrantDelegatedAdminRequest { Upn = Upn, TenantId = TenantA, Role = Constants.DelegatedRoles.DelegatedAdmin },
out var role));
Assert.Equal(Constants.DelegatedRoles.DelegatedAdmin, role);
}
// --- Revoke: 404-on-missing-row vs. audit + disconnect ---
[Fact]
public async Task RevokeCore_MissingRow_NoAuditNoDisconnect()
{
var (fn, repo, audit, signalR) = BuildAdminFn();
repo.Setup(r => r.RemoveDelegatedAdminAsync(UpnLower, TenantA)).ReturnsAsync(false);
var removed = await fn.RevokeCoreAsync(Upn, TenantA, "ga@vendor.example");
Assert.False(removed);
VerifyNoAudit(audit);
Assert.Empty(signalR.DisconnectedUsers);
}
[Fact]
public async Task RevokeCore_RemovedRow_AuditsAndDisconnectsLowercasedUpn()
{
var (fn, repo, audit, signalR) = BuildAdminFn();
repo.Setup(r => r.RemoveDelegatedAdminAsync(UpnLower, TenantA)).ReturnsAsync(true);
var removed = await fn.RevokeCoreAsync(Upn, TenantA, "ga@vendor.example");
Assert.True(removed);
// SignalR user ids are lowercased UPNs — a mixed-case route param must still hit the user's connections.
Assert.Equal(new[] { UpnLower }, signalR.DisconnectedUsers);
audit.Verify(a => a.LogAuditEntryAsync(
TenantA, "DELETE", "DelegatedAdmin", UpnLower, "ga@vendor.example", null), Times.Once);
}
// --- Disable is a revocation too; enable never disconnects ---
[Fact]
public async Task SetEnabledCore_MissingRow_NoAuditNoDisconnect()
{
var (fn, repo, audit, signalR) = BuildAdminFn();
repo.Setup(r => r.SetDelegatedAdminEnabledAsync(UpnLower, TenantA, false)).ReturnsAsync(false);
var ok = await fn.SetEnabledCoreAsync(Upn, TenantA, isEnabled: false, "ga@vendor.example");
Assert.False(ok);
VerifyNoAudit(audit);
Assert.Empty(signalR.DisconnectedUsers);
}
[Fact]
public async Task SetEnabledCore_Disable_Disconnects()
{
var (fn, repo, _, signalR) = BuildAdminFn();
repo.Setup(r => r.SetDelegatedAdminEnabledAsync(UpnLower, TenantA, false)).ReturnsAsync(true);
var ok = await fn.SetEnabledCoreAsync(Upn, TenantA, isEnabled: false, "ga@vendor.example");
Assert.True(ok);
Assert.Equal(new[] { UpnLower }, signalR.DisconnectedUsers);
}
[Fact]
public async Task SetEnabledCore_Enable_DoesNotDisconnect()
{
var (fn, repo, _, signalR) = BuildAdminFn();
repo.Setup(r => r.SetDelegatedAdminEnabledAsync(UpnLower, TenantA, true)).ReturnsAsync(true);
var ok = await fn.SetEnabledCoreAsync(Upn, TenantA, isEnabled: true, "ga@vendor.example");
Assert.True(ok);
Assert.Empty(signalR.DisconnectedUsers);
}
// --- Group unassign: 404-on-not-assigned vs. disconnect + per-tenant audit ---
[Fact]
public async Task UnassignCore_NotAssigned_NoAuditNoDisconnect()
{
var (fn, _, audit, signalR) = BuildGroupFn(); // default: no assignments for the UPN
var unassigned = await fn.UnassignCoreAsync(GroupId, Upn, "ga@vendor.example");
Assert.False(unassigned);
VerifyNoAudit(audit);
Assert.Empty(signalR.DisconnectedUsers);
}
[Fact]
public async Task UnassignCore_Assigned_DisconnectsAndAuditsPerGroupTenant()
{
var (fn, repo, audit, signalR) = BuildGroupFn();
repo.Setup(r => r.GetTenantGroupAsync(GroupId)).ReturnsAsync(new TenantGroup
{
GroupId = GroupId,
Name = "MSP Customers",
TenantIds = new List<string> { TenantA, TenantB },
AssigneeCount = 1,
});
repo.Setup(r => r.GetGroupAssignmentsForUpnAsync(UpnLower))
.ReturnsAsync(new List<TenantGroupAssignment> { new() { Upn = UpnLower, GroupId = GroupId } });
repo.Setup(r => r.UnassignGroupAsync(UpnLower, GroupId)).ReturnsAsync(true);
var unassigned = await fn.UnassignCoreAsync(GroupId, Upn, "ga@vendor.example");
Assert.True(unassigned);
Assert.Equal(new[] { UpnLower }, signalR.DisconnectedUsers);
// Lost access is audited under EACH of the group's tenants (customer-visible trail).
audit.Verify(a => a.LogAuditEntryAsync(
TenantA, "DELETE", "DelegatedGroupAccess", UpnLower, "ga@vendor.example",
It.IsAny<Dictionary<string, string>?>()), Times.Once);
audit.Verify(a => a.LogAuditEntryAsync(
TenantB, "DELETE", "DelegatedGroupAccess", UpnLower, "ga@vendor.example",
It.IsAny<Dictionary<string, string>?>()), Times.Once);
}
// --- Group delete: every (former) assignee's streams are cut ---
[Fact]
public async Task DeleteGroupCore_DisconnectsEveryAssignee_AndAuditsPerTenant()
{
var (fn, repo, audit, signalR) = BuildGroupFn();
var assignees = new List<TenantGroupAssignment>
{
new() { Upn = "one@partner.example", GroupId = GroupId },
new() { Upn = "two@partner.example", GroupId = GroupId },
};
repo.Setup(r => r.GetTenantGroupAsync(GroupId)).ReturnsAsync(new TenantGroup
{
GroupId = GroupId,
Name = "MSP Customers",
TenantIds = new List<string> { TenantA },
AssigneeCount = assignees.Count,
});
repo.Setup(r => r.GetGroupAssigneesAsync(GroupId)).ReturnsAsync(assignees);
repo.Setup(r => r.DeleteTenantGroupAsync(GroupId)).ReturnsAsync(true);
await fn.DeleteGroupCoreAsync(GroupId, "ga@vendor.example");
Assert.Equal(new[] { "one@partner.example", "two@partner.example" }, signalR.DisconnectedUsers);
audit.Verify(a => a.LogAuditEntryAsync(
TenantA, "DELETE", "DelegatedGroupAccess", "*", "ga@vendor.example",
It.IsAny<Dictionary<string, string>?>()), Times.Once);
}
}
@@ -3,6 +3,7 @@ using System.Collections.Generic;
using System.Linq;
using System.Threading.Tasks;
using Azure.Data.Tables;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Pagination;
using AutopilotMonitor.Functions.Services;
using AutopilotMonitor.Shared;
@@ -43,7 +44,9 @@ namespace AutopilotMonitor.Functions.DataAccess.TableStorage
["UserId"] = Truncate(entry.UserId, 128),
["Message"] = Truncate(entry.Message, 512),
["Details"] = Truncate(entry.Details, 4096),
["Timestamp"] = entry.Timestamp,
// "Timestamp" is a reserved system property (supplied values are ignored and
// any row rewrite resets it), so the event time needs its own column.
[BusinessTimestamp.OccurredUtcColumn] = entry.Timestamp,
};
await _table.UpsertEntityAsync(entity);
@@ -168,15 +171,16 @@ namespace AutopilotMonitor.Functions.DataAccess.TableStorage
if (!string.IsNullOrEmpty(lastRowKey))
clauses.Add($"RowKey gt '{lastRowKey!.Replace("'", "''")}'");
if (dateFrom.HasValue)
clauses.Add($"Timestamp ge datetime'{ToUtc(dateFrom.Value):o}'");
clauses.Add(BusinessTimestamp.OpsDateFromClause(ToUtc(dateFrom.Value)));
if (dateTo.HasValue)
clauses.Add($"Timestamp le datetime'{ToUtc(dateTo.Value):o}'");
clauses.Add(BusinessTimestamp.OpsDateToClause(ToUtc(dateTo.Value)));
return string.Join(" and ", clauses);
}
// Filters on the system-managed Timestamp; user-defined "Timestamp" property
// mirrors it on insert via SaveOpsEventAsync.
private static string? BuildFilter(string? category, DateTime? dateFrom, DateTime? dateTo)
// Date windows filter on the RowKey (reverse-tick, index-backed): the system
// Timestamp is reset by storage migrations, and an OccurredUtc property filter
// would exclude rows written before that column existed.
internal static string? BuildFilter(string? category, DateTime? dateFrom, DateTime? dateTo)
{
var clauses = new List<string>();
if (!string.IsNullOrEmpty(category))
@@ -185,11 +189,11 @@ namespace AutopilotMonitor.Functions.DataAccess.TableStorage
}
if (dateFrom.HasValue)
{
clauses.Add($"Timestamp ge datetime'{ToUtc(dateFrom.Value):o}'");
clauses.Add(BusinessTimestamp.OpsDateFromClause(ToUtc(dateFrom.Value)));
}
if (dateTo.HasValue)
{
clauses.Add($"Timestamp le datetime'{ToUtc(dateTo.Value):o}'");
clauses.Add(BusinessTimestamp.OpsDateToClause(ToUtc(dateTo.Value)));
}
return clauses.Count == 0 ? null : string.Join(" and ", clauses);
}
@@ -201,27 +205,30 @@ namespace AutopilotMonitor.Functions.DataAccess.TableStorage
{
var deleted = 0;
await foreach (var entity in _table.QueryAsync<TableEntity>())
// Server-side RowKey range instead of a full-table scan over the system
// Timestamp: every ops RowKey is a reverse-tick, so age is index-derivable,
// and the system Timestamp lies after storage migrations (it would freeze
// deletions for one retention period, then drop the whole pre-migration
// corpus in a single run).
var filter = BusinessTimestamp.OpsRetentionClause(cutoff);
await foreach (var entity in _table.QueryAsync<TableEntity>(
filter: filter, select: new[] { "PartitionKey", "RowKey" }))
{
var timestamp = entity.GetDateTimeOffset("Timestamp")?.UtcDateTime ?? DateTime.MinValue;
if (timestamp < cutoff)
try
{
try
{
await _table.DeleteEntityAsync(entity.PartitionKey, entity.RowKey);
deleted++;
}
catch (Exception ex)
{
_logger.LogWarning(ex, "Failed to delete ops event {PK}/{RK}", entity.PartitionKey, entity.RowKey);
}
await _table.DeleteEntityAsync(entity.PartitionKey, entity.RowKey);
deleted++;
}
catch (Exception ex)
{
_logger.LogWarning(ex, "Failed to delete ops event {PK}/{RK}", entity.PartitionKey, entity.RowKey);
}
}
return deleted;
}
private static OpsEventEntry MapToEntry(TableEntity entity)
internal static OpsEventEntry MapToEntry(TableEntity entity)
{
return new OpsEventEntry
{
@@ -233,10 +240,22 @@ namespace AutopilotMonitor.Functions.DataAccess.TableStorage
UserId = entity.GetString("UserId"),
Message = entity.GetString("Message") ?? string.Empty,
Details = entity.GetString("Details"),
Timestamp = entity.GetDateTimeOffset("Timestamp")?.UtcDateTime ?? DateTime.MinValue,
Timestamp = ResolveTimestamp(entity),
};
}
// Business time: OccurredUtc column → reverse-tick RowKey decode → system
// Timestamp (last resort; a row rewrite — e.g. storage migration — resets it).
internal static DateTime ResolveTimestamp(TableEntity entity)
{
var occurred = BusinessTimestamp.GetUtcDateTime(entity, BusinessTimestamp.OccurredUtcColumn);
if (occurred.HasValue)
return occurred.Value;
if (BusinessTimestamp.TryDecodeOpsRowKey(entity.RowKey, out var decoded))
return decoded;
return BusinessTimestamp.GetUtcDateTime(entity, "Timestamp") ?? DateTime.MinValue;
}
private static string? Truncate(string? value, int maxLength)
{
if (value == null) return null;
@@ -0,0 +1,103 @@
using System.Net;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Services;
using Microsoft.Azure.Functions.Worker;
using Microsoft.Azure.Functions.Worker.Http;
using Microsoft.Extensions.Logging;
namespace AutopilotMonitor.Functions.Functions.Admin
{
/// <summary>
/// Global-Admin trigger for the one-time OccurredUtc backfill on AuditLogs / OpsEvents
/// (business timestamps lost to the system-Timestamp reset of the 2026-07-18 storage
/// migration; values are recovered from the reverse-tick RowKeys). DRY-RUN BY DEFAULT —
/// a real run requires the explicit <c>dryRun=false</c> query parameter, so the operator
/// always sees counts and samples before anything is written.
/// </summary>
public class BackfillOccurredUtcFunction
{
private readonly ILogger<BackfillOccurredUtcFunction> _logger;
private readonly OccurredUtcBackfillService _service;
private const int DefaultMaxRows = 250;
private const int MaxMaxRows = 1000;
public BackfillOccurredUtcFunction(
ILogger<BackfillOccurredUtcFunction> logger,
OccurredUtcBackfillService service)
{
_logger = logger;
_service = service;
}
/// <summary>
/// POST /api/maintenance/backfill-occurred-utc
/// Query parameters:
/// - table: "audit" | "ops" (required).
/// - dryRun: "false" to actually write; anything else (or absent) reports only.
/// - maxRows: rows examined per invocation (default 250, max 1000). Follow
/// nextContinuation to continue; null nextContinuation = table done.
/// - continuation: opaque token from the previous response (URL-encode it).
/// </summary>
[Function("BackfillOccurredUtc")]
public async Task<HttpResponseData> Run(
[HttpTrigger(AuthorizationLevel.Anonymous, "post", Route = "maintenance/backfill-occurred-utc")] HttpRequestData req,
FunctionContext context)
{
// Authentication + GlobalAdminOnly authorization enforced by PolicyEnforcementMiddleware
var userEmail = TenantHelper.GetUserIdentifier(req);
var table = (req.Query["table"] ?? string.Empty).Trim().ToLowerInvariant();
if (table != OccurredUtcBackfillService.TableAudit && table != OccurredUtcBackfillService.TableOps)
{
var bad = req.CreateResponse(HttpStatusCode.BadRequest);
await bad.WriteAsJsonAsync(new
{
error = $"table must be '{OccurredUtcBackfillService.TableAudit}' or '{OccurredUtcBackfillService.TableOps}'"
});
return bad;
}
var dryRun = !string.Equals(req.Query["dryRun"], "false", StringComparison.OrdinalIgnoreCase);
var continuation = req.Query["continuation"];
var maxRows = DefaultMaxRows;
if (!string.IsNullOrEmpty(req.Query["maxRows"]))
{
if (!int.TryParse(req.Query["maxRows"], out maxRows) || maxRows < 1)
{
var bad = req.CreateResponse(HttpStatusCode.BadRequest);
await bad.WriteAsJsonAsync(new { error = $"maxRows must be a positive integer (max {MaxMaxRows})" });
return bad;
}
maxRows = Math.Min(maxRows, MaxMaxRows);
}
_logger.LogInformation(
"OccurredUtc backfill requested by {User}: table={Table} dryRun={DryRun} maxRows={MaxRows} hasContinuation={HasContinuation}",
userEmail, table, dryRun, maxRows, !string.IsNullOrEmpty(continuation));
try
{
var result = await _service.RunAsync(table, dryRun, maxRows, continuation);
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new
{
success = true,
result,
triggeredBy = userEmail,
triggeredAt = DateTime.UtcNow,
});
return response;
}
catch (Exception ex)
{
_logger.LogError(ex, "OccurredUtc backfill run failed");
var error = req.CreateResponse(HttpStatusCode.InternalServerError);
await error.WriteAsJsonAsync(new { success = false, message = "Internal server error" });
return error;
}
}
}
}
@@ -64,19 +64,9 @@ public class DelegatedAdminManagementFunction
var currentUpn = context.GetRequestContext().UserPrincipalName;
var body = await req.ReadFromJsonAsync<GrantDelegatedAdminRequest>();
if (body == null || string.IsNullOrWhiteSpace(body.Upn) || string.IsNullOrWhiteSpace(body.TenantId))
return await Bad(req, "upn and tenantId are required");
// Mirror the tenant-group path: a mistyped tenantId must not create a garbage scope entry
// that pollutes the allow-list and seeds a per-tenant audit trail keyed on the typo.
if (!Guid.TryParse(body.TenantId, out _))
return await Bad(req, "a valid tenantId (GUID) is required");
// Default to the least-privileged role; reject anything we don't recognize (fail-closed) so a typo
// can never silently widen access.
var role = string.IsNullOrWhiteSpace(body.Role) ? Constants.DelegatedRoles.DelegatedReader : body.Role;
if (role != Constants.DelegatedRoles.DelegatedReader && role != Constants.DelegatedRoles.DelegatedAdmin)
return await Bad(req, $"role must be '{Constants.DelegatedRoles.DelegatedReader}' or '{Constants.DelegatedRoles.DelegatedAdmin}'");
var validationError = ValidateGrantRequest(body, out var role);
if (validationError != null)
return await Bad(req, validationError);
// Edition note: TARGET tenants may be any edition — an MSP on Enterprise may manage Community
// customers. The Enterprise requirement applies to the delegated admin's HOME tenant and is
@@ -84,7 +74,7 @@ public class DelegatedAdminManagementFunction
// be checked here: at grant time only the UPN is known, and UPN-domain → tenant mapping is not
// reliable (multi-domain tenants). Grants to non-Enterprise-homed admins are simply inert.
var entry = await _delegatedAdminService.UpsertAsync(
body.Upn, body.TenantId, role,
body!.Upn, body.TenantId, role,
Constants.DelegatedStatus.Active, Constants.DelegatedSource.OperatorGranted, currentUpn ?? "");
await _maintenanceRepo.LogAuditEntryAsync(
@@ -112,10 +102,7 @@ public class DelegatedAdminManagementFunction
string upn, string tenantId, FunctionContext context)
{
var currentUpn = context.GetRequestContext().UserPrincipalName;
// The service reports whether a row was actually deleted — a mistyped upn/tenantId must not
// 200 and write a false customer-visible "access removed" audit row while the real grant
// stays live (mirrors the tenant-group revoke semantics).
var removed = await _delegatedAdminService.RemoveAsync(upn, tenantId);
var removed = await RevokeCoreAsync(upn, tenantId, currentUpn);
if (!removed)
{
var notFound = req.CreateResponse(HttpStatusCode.NotFound);
@@ -123,20 +110,34 @@ public class DelegatedAdminManagementFunction
return notFound;
}
await _maintenanceRepo.LogAuditEntryAsync(
tenantId.ToLowerInvariant(), "DELETE", "DelegatedAdmin", upn.ToLowerInvariant(), currentUpn ?? "");
// Enforcement: cut the revoked caller's already-joined live streams — group authorization is
// join-time only, so without this they keep receiving tenant telemetry until the connection drops.
await _signalRService.DisconnectUserAsync(upn.ToLowerInvariant());
_logger.LogInformation("Delegated revoke: {Upn} -> {TenantId} by {By}", upn, tenantId, currentUpn);
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = "Delegated assignment revoked" });
return response;
}
/// <summary>
/// Testable core of the revoke flow (no HTTP plumbing — see GetAllBlockedDevicesFunctionTests for the
/// convention). Returns false when no assignment row existed (caller 404s): a mistyped upn/tenantId
/// must not 200 and write a false customer-visible "access removed" audit row while the real grant
/// stays live (mirrors the tenant-group revoke semantics). On a real removal writes the audit entry
/// and cuts the revoked UPN's live streams — group authorization is join-time only, so without the
/// kick they keep receiving tenant telemetry until the connection drops.
/// </summary>
internal async Task<bool> RevokeCoreAsync(string upn, string tenantId, string? currentUpn)
{
var removed = await _delegatedAdminService.RemoveAsync(upn, tenantId);
if (!removed)
return false;
await _maintenanceRepo.LogAuditEntryAsync(
tenantId.ToLowerInvariant(), "DELETE", "DelegatedAdmin", upn.ToLowerInvariant(), currentUpn ?? "");
await _signalRService.DisconnectUserAsync(upn.ToLowerInvariant());
_logger.LogInformation("Delegated revoke: {Upn} -> {TenantId} by {By}", upn, tenantId, currentUpn);
return true;
}
/// <summary>PATCH /api/global/delegated-admins/{upn}/{tenantId}/enable — re-enable a disabled assignment. GlobalAdminOnly.</summary>
[Function("EnableDelegatedAdmin")]
[Authorize]
@@ -157,7 +158,7 @@ public class DelegatedAdminManagementFunction
HttpRequestData req, string upn, string tenantId, bool isEnabled, FunctionContext context)
{
var currentUpn = context.GetRequestContext().UserPrincipalName;
var ok = await _delegatedAdminService.SetEnabledAsync(upn, tenantId, isEnabled);
var ok = await SetEnabledCoreAsync(upn, tenantId, isEnabled, currentUpn);
if (!ok)
{
var notFound = req.CreateResponse(HttpStatusCode.NotFound);
@@ -165,20 +166,56 @@ public class DelegatedAdminManagementFunction
return notFound;
}
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = isEnabled ? "Delegated assignment enabled" : "Delegated assignment disabled" });
return response;
}
/// <summary>
/// Testable core of enable/disable (no HTTP plumbing). Returns false when no assignment row
/// existed (caller 404s — no audit, no disconnect). Disable is a revocation too, so it cuts the
/// UPN's already-joined live streams (see <see cref="RevokeCoreAsync"/>); enable never disconnects.
/// </summary>
internal async Task<bool> SetEnabledCoreAsync(string upn, string tenantId, bool isEnabled, string? currentUpn)
{
var ok = await _delegatedAdminService.SetEnabledAsync(upn, tenantId, isEnabled);
if (!ok)
return false;
await _maintenanceRepo.LogAuditEntryAsync(
tenantId.ToLowerInvariant(), "UPDATE", "DelegatedAdmin", upn.ToLowerInvariant(), currentUpn ?? "",
new Dictionary<string, string> { { "IsEnabled", isEnabled.ToString() } });
// Disable is a revocation too — cut already-joined live streams (see RevokeDelegatedAdmin).
if (!isEnabled)
await _signalRService.DisconnectUserAsync(upn.ToLowerInvariant());
_logger.LogInformation("Delegated {Action}: {Upn} -> {TenantId} by {By}",
isEnabled ? "enable" : "disable", upn, tenantId, currentUpn);
return true;
}
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = isEnabled ? "Delegated assignment enabled" : "Delegated assignment disabled" });
return response;
/// <summary>
/// Pure grant-request validation (testable seam). Returns the 400 error message, or null when valid —
/// then <paramref name="role"/> carries the effective role. A mistyped tenantId must not create a
/// garbage scope entry that pollutes the allow-list and seeds a per-tenant audit trail keyed on the
/// typo; the role defaults to least-privileged and unknown roles are rejected (fail-closed) so a typo
/// can never silently widen access.
/// </summary>
internal static string? ValidateGrantRequest(GrantDelegatedAdminRequest? body, out string role)
{
role = Constants.DelegatedRoles.DelegatedReader;
if (body == null || string.IsNullOrWhiteSpace(body.Upn) || string.IsNullOrWhiteSpace(body.TenantId))
return "upn and tenantId are required";
if (!Guid.TryParse(body.TenantId, out _))
return "a valid tenantId (GUID) is required";
role = string.IsNullOrWhiteSpace(body.Role) ? Constants.DelegatedRoles.DelegatedReader : body.Role!;
if (role != Constants.DelegatedRoles.DelegatedReader && role != Constants.DelegatedRoles.DelegatedAdmin)
return $"role must be '{Constants.DelegatedRoles.DelegatedReader}' or '{Constants.DelegatedRoles.DelegatedAdmin}'";
return null;
}
private static async Task<HttpResponseData> Bad(HttpRequestData req, string error)
@@ -112,15 +112,26 @@ public class TenantGroupManagementFunction
string groupId, FunctionContext context)
{
var currentUpn = context.GetRequestContext().UserPrincipalName;
await DeleteGroupCoreAsync(groupId, currentUpn);
// Capture tenants + assignees BEFORE the cascade delete removes them, so we can audit the
// bulk access removal under each affected tenant and cut every assignee's live streams.
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = "Group deleted" });
return response;
}
/// <summary>
/// Testable core of the group cascade delete (no HTTP plumbing). Captures tenants + assignees
/// BEFORE the delete removes them, so the bulk access removal is audited under each affected
/// tenant and EVERY (former) assignee's live streams are cut — group authorization is join-time
/// only, so without the kick they keep receiving the managed tenants' telemetry until their
/// connection drops.
/// </summary>
internal async Task DeleteGroupCoreAsync(string groupId, string? currentUpn)
{
var group = await _delegatedAdminService.GetGroupAsync(groupId);
var assignees = await _delegatedAdminService.GetGroupAssigneesAsync(groupId);
await _delegatedAdminService.DeleteGroupAsync(groupId);
// Enforcement: group authorization is join-time only — without the kick, every (former)
// assignee keeps receiving the managed tenants' live telemetry until their connection drops.
foreach (var assignee in assignees)
await _signalRService.DisconnectUserAsync(assignee.Upn);
@@ -137,10 +148,6 @@ public class TenantGroupManagementFunction
}
_logger.LogInformation("Tenant group deleted: {GroupId} by {By}", groupId, currentUpn);
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = "Group deleted" });
return response;
}
/// <summary>POST /api/global/tenant-groups/{groupId}/tenants — add a tenant. GlobalAdminOnly. Body: { "tenantId": "&lt;guid&gt;" }.</summary>
@@ -271,17 +278,31 @@ public class TenantGroupManagementFunction
string groupId, string upn, FunctionContext context)
{
var currentUpn = context.GetRequestContext().UserPrincipalName;
var unassigned = await UnassignCoreAsync(groupId, upn, currentUpn);
if (!unassigned)
return await NotFound(req);
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = "Unassigned from group" });
return response;
}
/// <summary>
/// Testable core of the unassign flow (no HTTP plumbing). Returns false when the UPN was not
/// actually assigned (caller 404s) — so a mistyped UPN can't 200 and write false "access removed"
/// audit rows, and no disconnect fires. On a real unassign cuts the UPN's already-joined live
/// streams (join-time-only authz) and audits the lost access under each of the group's tenants.
/// </summary>
internal async Task<bool> UnassignCoreAsync(string groupId, string upn, string? currentUpn)
{
var normalizedUpn = upn.ToLowerInvariant();
// Read the group's tenants (unchanged by unassign) so we can audit the UPN's lost access per tenant.
var group = await _delegatedAdminService.GetGroupAsync(groupId);
// The service returns false when the UPN was not actually assigned — so a mistyped UPN can't 200 and
// write false "access removed" audit rows. Only audit after a real unassign.
var unassigned = await _delegatedAdminService.UnassignGroupAsync(normalizedUpn, groupId);
if (!unassigned)
return await NotFound(req);
return false;
// Enforcement: cut the unassigned caller's already-joined live streams (join-time-only authz).
await _signalRService.DisconnectUserAsync(normalizedUpn);
if (group != null)
@@ -296,10 +317,7 @@ public class TenantGroupManagementFunction
}
_logger.LogInformation("Group {GroupId} unassigned from {Upn} by {By}", groupId, normalizedUpn, currentUpn);
var response = req.CreateResponse(HttpStatusCode.OK);
await response.WriteAsJsonAsync(new { message = "Unassigned from group" });
return response;
return true;
}
/// <summary>Logs one audit entry per tenant (the customer-visible "who can read my tenant" trail).</summary>
@@ -10,6 +10,7 @@ using AutopilotMonitor.Functions.Services;
using AutopilotMonitor.Shared.Models;
using AutopilotMonitor.Shared.Pagination;
using Microsoft.Azure.Functions.Worker;
using Microsoft.Extensions.Caching.Memory;
using Microsoft.Azure.Functions.Worker.Http;
using Microsoft.Extensions.Logging;
@@ -17,15 +18,23 @@ namespace AutopilotMonitor.Functions.Functions.Config
{
public class GetAllTenantConfigurationsFunction
{
// TTL for the "this managed id has NO config row in any casing" verdict a rescue scan proved.
// Bounds how long a tenant onboarded moments after that scan stays absent from a delegated
// caller's config/all (matches the positive config cache TTL in TenantConfigurationService).
private static readonly TimeSpan MissingConfigNegativeCacheDuration = TimeSpan.FromMinutes(5);
private readonly ILogger<GetAllTenantConfigurationsFunction> _logger;
private readonly TenantConfigurationService _configService;
private readonly IMemoryCache _cache;
public GetAllTenantConfigurationsFunction(
ILogger<GetAllTenantConfigurationsFunction> logger,
TenantConfigurationService configService)
TenantConfigurationService configService,
IMemoryCache cache)
{
_logger = logger;
_configService = configService;
_cache = cache;
}
[Function("GetAllTenantConfigurations")]
@@ -73,9 +82,21 @@ namespace AutopilotMonitor.Functions.Functions.Config
// (agent paths auto-create rows with the caller-supplied id verbatim) — the same caveat
// the delegated sessions fan-out handles case-insensitively. A point-read is exact-case,
// so any id it missed gets ONE scan+filter rescue pass (the pre-point-read behavior).
// Only stale grants or case-variant rows pay for it; the common path stays read-bounded.
// An id a previous scan already proved missing in ANY casing is negative-cached and
// skipped for the TTL — otherwise a single stale grant (offboarded / never-onboarded
// tenant) would make EVERY config/all call pay a full table scan.
if (subset.Count < managedIds.Count)
subset = MergeRescuedConfigs(managedIds, subset, await _configService.GetAllConfigurationsAsync());
{
var unproven = MissingManagedIds(managedIds, subset)
.Where(id => !_cache.TryGetValue(MissingConfigCacheKey(id), out _))
.ToList();
if (unproven.Count > 0)
{
subset = MergeRescuedConfigs(managedIds, subset, await _configService.GetAllConfigurationsAsync());
foreach (var id in MissingManagedIds(unproven, subset))
_cache.Set(MissingConfigCacheKey(id), true, MissingConfigNegativeCacheDuration);
}
}
_logger.LogInformation("GetAllTenantConfigurations (delegated subset, {Count} tenants) by {User}",
subset.Count, userIdentifier);
@@ -177,6 +198,23 @@ namespace AutopilotMonitor.Functions.Functions.Config
IEnumerable<TenantConfiguration?> reads)
=> reads.Where(c => c != null).Select(c => c!).ToList();
/// <summary>
/// The managed ids not present (case-insensitively) in <paramref name="configs"/> — before the
/// rescue scan these are the point-read misses, after it the proven-missing ids to negative-cache.
/// Pure + testable.
/// </summary>
internal static List<string> MissingManagedIds(
IEnumerable<string> managedIds, IReadOnlyCollection<TenantConfiguration> configs)
{
var have = new HashSet<string>(configs.Select(c => c.TenantId), StringComparer.OrdinalIgnoreCase);
return managedIds.Where(id => !have.Contains(id)).ToList();
}
/// <summary>Negative-cache key for a managed id a rescue scan proved has no config row.
/// Lowercased so case variants of the same grant share one verdict.</summary>
internal static string MissingConfigCacheKey(string tenantId)
=> $"tenant-config-missing:{tenantId.ToLowerInvariant()}";
/// <summary>
/// Case-variant rescue for managed ids the exact-case point-reads missed: from the full config scan,
/// add every config whose TenantId matches a managed id case-insensitively and is not already in the
@@ -0,0 +1,154 @@
using System;
using System.Globalization;
using Azure.Data.Tables;
namespace AutopilotMonitor.Functions.Helpers
{
/// <summary>
/// Business-timestamp handling for tables whose rows historically relied on the
/// Azure Tables system <c>Timestamp</c> (= row write time). That system property is
/// server-managed: supplied values are ignored on write and every row rewrite
/// (storage migration, merge) resets it. The authoritative event time therefore
/// lives in (a) the explicit <see cref="OccurredUtcColumn"/> custom property on
/// rows written after its introduction, and (b) the time encoded in the RowKey
/// for older rows. Readers resolve OccurredUtc → RowKey decode → system Timestamp.
///
/// RowKey schemes decoded here:
/// AuditLogs: "!{MaxTicks - ticks:D19}_{guid:N}" (legacy pre-2026-05 rows are bare GUIDs — not decodable)
/// OpsEvents: "{MaxTicks - ticks:D19}"
/// Events: "{timestamp:yyyyMMddHHmmssfff}_{sequence:D10}"
/// </summary>
internal static class BusinessTimestamp
{
/// <summary>Custom column carrying the business event time. Survives migrations verbatim
/// (unlike the system Timestamp, which any row rewrite resets).</summary>
internal const string OccurredUtcColumn = "OccurredUtc";
internal const string AuditRowKeyPrefix = "!";
/// <summary>
/// '"' (0x22) is the character immediately after '!' (0x21). `RowKey lt '"'` confines a
/// range to time-encoded audit rows: legacy bare-GUID RowKeys start with a hex digit
/// (≥ '0', 0x30) and would otherwise match ANY lower bound of the form `RowKey ge '!…'`.
/// </summary>
internal const string AuditTimeEncodedUpperBound = "\"";
private const string EventRowKeyTimestampFormat = "yyyyMMddHHmmssfff";
/// <summary>
/// Reads a datetime column as UTC regardless of the SDK's materialized type —
/// service reads surface Edm.DateTime as either DateTimeOffset or DateTime
/// depending on payload shape (the pre-existing mappers already read both).
/// Unspecified kinds are treated as UTC (all stored values are UTC).
/// </summary>
internal static DateTime? GetUtcDateTime(TableEntity entity, string column)
{
if (!entity.TryGetValue(column, out var value) || value == null)
return null;
return value switch
{
DateTimeOffset dto => dto.UtcDateTime,
DateTime dt => dt.Kind == DateTimeKind.Local
? dt.ToUniversalTime()
: DateTime.SpecifyKind(dt, DateTimeKind.Utc),
_ => null,
};
}
// ===== RowKey decoders =====
/// <summary>Decodes an audit RowKey ("!{rev:D19}_{guid:N}") to the original UTC write time.
/// Returns false for legacy bare-GUID RowKeys and any other foreign format.</summary>
internal static bool TryDecodeAuditRowKey(string? rowKey, out DateTime utc)
{
utc = default;
if (rowKey == null || rowKey.Length < 21 || rowKey[0] != '!')
return false;
return TryDecodeReverseTicks(rowKey.AsSpan(1, 19), out utc);
}
/// <summary>Decodes an ops-event RowKey (bare "{rev:D19}") to the original UTC write time.</summary>
internal static bool TryDecodeOpsRowKey(string? rowKey, out DateTime utc)
{
utc = default;
if (rowKey == null || rowKey.Length != 19)
return false;
return TryDecodeReverseTicks(rowKey.AsSpan(), out utc);
}
/// <summary>Decodes an enrollment-event RowKey ("{yyyyMMddHHmmssfff}_{seq:D10}") to the
/// (sanitized) agent event time, millisecond precision, Kind=Utc.</summary>
internal static bool TryDecodeEventRowKeyPrefix(string? rowKey, out DateTime utc)
{
utc = default;
if (rowKey == null)
return false;
var separator = rowKey.IndexOf('_');
if (separator != 17)
return false;
if (!DateTime.TryParseExact(rowKey.Substring(0, 17), EventRowKeyTimestampFormat,
CultureInfo.InvariantCulture,
DateTimeStyles.AssumeUniversal | DateTimeStyles.AdjustToUniversal, out utc))
return false;
return true;
}
private static bool TryDecodeReverseTicks(ReadOnlySpan<char> revTickDigits, out DateTime utc)
{
utc = default;
foreach (var c in revTickDigits)
{
if (c < '0' || c > '9')
return false;
}
if (!long.TryParse(revTickDigits, NumberStyles.None, CultureInfo.InvariantCulture, out var rev))
return false;
var ticks = DateTime.MaxValue.Ticks - rev;
if (ticks < DateTime.MinValue.Ticks || ticks > DateTime.MaxValue.Ticks)
return false;
utc = new DateTime(ticks, DateTimeKind.Utc);
return true;
}
// ===== OData date-window clauses (RowKey ranges) =====
// Date windows filter on the RowKey rather than a property: rows missing a property are
// excluded from OData comparisons (so an OccurredUtc filter would drop every not-yet-
// backfilled row), and RowKey ranges are index-backed. String order equals numeric order
// because the reverse-tick is fixed-width D19. rev(t) = MaxTicks - t.Ticks, so larger
// timestamps have lexically smaller RowKeys (newest-first).
private static long Rev(DateTime utc) => DateTime.MaxValue.Ticks - utc.Ticks;
/// <summary>Audit rows with business time ≥ <paramref name="fromUtc"/> (inclusive):
/// rev ≤ rev(from) &lt; rev(from)+1. The exclusive upper bound also excludes all legacy
/// GUID rows (they sort after every '!' row).</summary>
internal static string AuditDateFromClause(DateTime fromUtc)
=> $"RowKey lt '{AuditRowKeyPrefix}{Rev(fromUtc) + 1:D19}'";
/// <summary>Audit rows with business time ≤ <paramref name="toUtc"/> (inclusive):
/// rev ≥ rev(to). The lower bound alone would match every legacy GUID row, hence the
/// mandatory time-encoded guard.</summary>
internal static string AuditDateToClause(DateTime toUtc)
=> $"RowKey ge '{AuditRowKeyPrefix}{Rev(toUtc):D19}' and RowKey lt '{AuditTimeEncodedUpperBound}'";
/// <summary>Audit rows strictly older than <paramref name="cutoffUtc"/> (retention):
/// rev ≥ rev(cutoff)+1 ⇔ ts &lt; cutoff. MUST keep the time-encoded guard — without it
/// this clause matches (and a retention sweep would delete) every legacy GUID row.</summary>
internal static string AuditRetentionClause(DateTime cutoffUtc)
=> $"RowKey ge '{AuditRowKeyPrefix}{Rev(cutoffUtc) + 1:D19}' and RowKey lt '{AuditTimeEncodedUpperBound}'";
/// <summary>Ops rows with business time ≥ <paramref name="fromUtc"/> (inclusive). Bare D19
/// RowKey (no suffix), so `le` is exact.</summary>
internal static string OpsDateFromClause(DateTime fromUtc)
=> $"RowKey le '{Rev(fromUtc):D19}'";
/// <summary>Ops rows with business time ≤ <paramref name="toUtc"/> (inclusive).</summary>
internal static string OpsDateToClause(DateTime toUtc)
=> $"RowKey ge '{Rev(toUtc):D19}'";
/// <summary>Ops rows strictly older than <paramref name="cutoffUtc"/> (retention):
/// `gt` not `ge` — a row at exactly the cutoff is kept, matching `Timestamp lt` semantics.</summary>
internal static string OpsRetentionClause(DateTime cutoffUtc)
=> $"RowKey gt '{Rev(cutoffUtc):D19}'";
}
}
@@ -252,6 +252,7 @@ builder.Services.AddHttpClient<GitHubRuleRepository>()
.AddPolicyHandler((sp, _) => sp.GetRequiredService<ResiliencePolicies>().ExternalDataApi);
builder.Services.AddSingleton<MaintenanceService>();
builder.Services.AddSingleton<LegacyReclassificationService>();
builder.Services.AddSingleton<OccurredUtcBackfillService>();
builder.Services.AddSingleton<
AutopilotMonitor.Functions.Services.Monitoring.IAzureMonitorMetricsReader,
AutopilotMonitor.Functions.Services.Monitoring.AzureMonitorMetricsReader>();
@@ -399,6 +399,9 @@ public static class EndpointAccessPolicyCatalog
// One-time retro-reconcile of misclassified historical sessions (audit 2026-07-16) —
// mutating, dry-run by default, GA only.
new("POST", "maintenance/reclassify-legacy", EndpointPolicy.GlobalAdminOnly),
// One-time OccurredUtc backfill on AuditLogs/OpsEvents (2026-07-18 migration reset
// the system Timestamps) — mutating, dry-run by default, GA only.
new("POST", "maintenance/backfill-occurred-utc", EndpointPolicy.GlobalAdminOnly),
// ── Critical-Table Backup (plan §PR1+§PR2) ─────────────────────────────
new("POST", "global/backups/trigger", EndpointPolicy.GlobalAdminOnly),
@@ -0,0 +1,169 @@
using Azure;
using Azure.Data.Tables;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Pagination;
using AutopilotMonitor.Shared;
using Microsoft.Extensions.Logging;
namespace AutopilotMonitor.Functions.Services
{
/// <summary>
/// One-time backfill of the <c>OccurredUtc</c> business-timestamp column on
/// AuditLogs and OpsEvents rows written before the column existed. The value is
/// decoded deterministically from the reverse-tick RowKey, so re-runs are
/// idempotent and the write is a pure MERGE that adds exactly one property
/// (nothing existing is modified or deleted). Audit legacy bare-GUID rows carry
/// no recoverable time and are excluded server-side by the scan filter.
/// Readers do not depend on this backfill (they decode the RowKey on the fly);
/// its value is future-migration robustness and raw-query readability.
/// </summary>
public class OccurredUtcBackfillService
{
public const string TableAudit = "audit";
public const string TableOps = "ops";
private const int MaxSamplesInResponse = 100;
private const int TransactionChunkSize = 100;
private readonly TableStorageService _storage;
private readonly ILogger<OccurredUtcBackfillService> _logger;
public OccurredUtcBackfillService(TableStorageService storage, ILogger<OccurredUtcBackfillService> logger)
{
_storage = storage;
_logger = logger;
}
public class BackfillResult
{
public string Table { get; set; } = string.Empty;
public bool DryRun { get; set; }
public int RowsExamined { get; set; }
public int WouldWrite { get; set; }
public int Written { get; set; }
public int SkippedAlreadySet { get; set; }
public int SkippedUndecodable { get; set; }
public int Errors { get; set; }
public string? NextContinuation { get; set; }
public List<BackfillSample> Samples { get; set; } = new();
}
public class BackfillSample
{
public string PartitionKey { get; set; } = string.Empty;
public string RowKey { get; set; } = string.Empty;
public DateTime DecodedUtc { get; set; }
}
internal enum RowDecision { AlreadySet, Write, Undecodable }
/// <summary>Pure per-row decision, extracted for unit testing.</summary>
internal static RowDecision DecideRow(string table, TableEntity entity, out DateTime occurredUtc)
{
occurredUtc = default;
if (BusinessTimestamp.GetUtcDateTime(entity, BusinessTimestamp.OccurredUtcColumn).HasValue)
return RowDecision.AlreadySet;
var decoded = table == TableAudit
? BusinessTimestamp.TryDecodeAuditRowKey(entity.RowKey, out occurredUtc)
: BusinessTimestamp.TryDecodeOpsRowKey(entity.RowKey, out occurredUtc);
return decoded ? RowDecision.Write : RowDecision.Undecodable;
}
public async Task<BackfillResult> RunAsync(string table, bool dryRun, int maxRows, string? continuation)
{
var (tableName, scanFilter) = table switch
{
// Audit: server-side restriction to time-encoded '!'-rows — legacy bare-GUID
// rows (which sort after '"') have nothing to decode and are never touched.
TableAudit => (Constants.TableNames.AuditLogs,
$"RowKey ge '{BusinessTimestamp.AuditRowKeyPrefix}' and RowKey lt '{BusinessTimestamp.AuditTimeEncodedUpperBound}'"),
TableOps => (Constants.TableNames.OpsEvents, (string?)null),
_ => throw new ArgumentException($"Unknown table '{table}'", nameof(table)),
};
var tableClient = _storage.GetTableClient(tableName);
var result = new BackfillResult { Table = table, DryRun = dryRun };
var (entities, nextToken) = await AzureTablesPaginator.FetchPageAsync<TableEntity>(
client: tableClient,
filter: scanFilter,
pageSize: maxRows,
continuation: continuation,
select: new[] { "PartitionKey", "RowKey", BusinessTimestamp.OccurredUtcColumn });
result.NextContinuation = nextToken;
var pending = new List<(string PartitionKey, string RowKey, DateTime OccurredUtc)>();
foreach (var entity in entities)
{
result.RowsExamined++;
switch (DecideRow(table, entity, out var occurredUtc))
{
case RowDecision.AlreadySet:
result.SkippedAlreadySet++;
break;
case RowDecision.Undecodable:
result.SkippedUndecodable++;
_logger.LogWarning("OccurredUtc backfill: undecodable RowKey {PK}/{RK} in {Table}",
entity.PartitionKey, entity.RowKey, tableName);
break;
case RowDecision.Write:
result.WouldWrite++;
pending.Add((entity.PartitionKey, entity.RowKey, occurredUtc));
if (result.Samples.Count < MaxSamplesInResponse)
{
result.Samples.Add(new BackfillSample
{
PartitionKey = entity.PartitionKey,
RowKey = entity.RowKey,
DecodedUtc = occurredUtc,
});
}
break;
}
}
if (!dryRun && pending.Count > 0)
result.Written = await WritePendingAsync(tableClient, pending, result);
_logger.LogInformation(
"OccurredUtc backfill {Table}: dryRun={DryRun} examined={Examined} wouldWrite={WouldWrite} written={Written} alreadySet={AlreadySet} undecodable={Undecodable} errors={Errors} more={More}",
table, dryRun, result.RowsExamined, result.WouldWrite, result.Written,
result.SkippedAlreadySet, result.SkippedUndecodable, result.Errors, result.NextContinuation != null);
return result;
}
private async Task<int> WritePendingAsync(TableClient tableClient,
List<(string PartitionKey, string RowKey, DateTime OccurredUtc)> pending, BackfillResult result)
{
var written = 0;
foreach (var partitionGroup in pending.GroupBy(p => p.PartitionKey))
{
foreach (var chunk in partitionGroup.Chunk(TransactionChunkSize))
{
// Merge entity carries ONLY the new column; unconditional (ETag *) merge is
// safe because the value is a pure function of the immutable RowKey.
var actions = chunk.Select(p => new TableTransactionAction(
TableTransactionActionType.UpdateMerge,
new TableEntity(p.PartitionKey, p.RowKey)
{
[BusinessTimestamp.OccurredUtcColumn] = p.OccurredUtc,
},
ETag.All)).ToList();
try
{
await tableClient.SubmitTransactionAsync(actions);
written += actions.Count;
}
catch (Exception ex)
{
result.Errors += actions.Count;
_logger.LogError(ex, "OccurredUtc backfill: merge transaction failed for partition {PK} ({Count} rows)",
partitionGroup.Key, actions.Count);
}
}
}
return written;
}
}
}
@@ -1,5 +1,6 @@
using Azure;
using Azure.Data.Tables;
using AutopilotMonitor.Functions.Helpers;
using AutopilotMonitor.Functions.Pagination;
using AutopilotMonitor.Functions.Security;
using AutopilotMonitor.Shared;
@@ -55,7 +56,10 @@ namespace AutopilotMonitor.Functions.Services
{ "EntityType", entityType },
{ "EntityId", entityId },
{ "PerformedBy", performedBy },
{ "Timestamp", timestamp },
// Business event time. "Timestamp" is a reserved system property (server
// ignores supplied values and resets it on every row rewrite), so the
// event time must live in a regular custom column to survive migrations.
{ BusinessTimestamp.OccurredUtcColumn, timestamp },
{ "Details", details != null ? JsonConvert.SerializeObject(details) : string.Empty }
};
@@ -71,32 +75,26 @@ namespace AutopilotMonitor.Functions.Services
}
/// <summary>
/// Retention cleanup: deletes audit log entries whose Timestamp is older than the cutoff.
/// The AuditLogs table is append-only (one row per admin action) and is otherwise only
/// wiped on tenant offboarding, so without this it grows unbounded. Filters server-side on
/// the per-row "Timestamp" property and selects only PK/RK to keep the scan cheap.
/// Retention cleanup: deletes audit log entries whose business time is older than the
/// cutoff. The AuditLogs table is append-only (one row per admin action) and is otherwise
/// only wiped on tenant offboarding, so without this it grows unbounded.
/// Two passes: (1) time-encoded '!'-rows via an index-backed RowKey range — the system
/// Timestamp cannot be used because storage migrations reset it (post-migration it would
/// delete nothing for a retention period, then the whole pre-migration corpus at once);
/// (2) legacy bare-GUID rows (true event time unrecoverable) via their frozen system
/// Timestamp — they all age out ~180d after the 2026-07-18 migration.
/// TODO: Remove pass 2 after 2027-03-01 (no bare-GUID rows can remain by then).
/// </summary>
public async Task<int> DeleteAuditLogsOlderThanAsync(DateTime cutoffUtc)
{
try
{
var tableClient = _tableServiceClient.GetTableClient(Constants.TableNames.AuditLogs);
var filter = $"Timestamp lt datetime'{cutoffUtc:yyyy-MM-ddTHH:mm:ss}Z'";
var query = tableClient.QueryAsync<TableEntity>(filter: filter, select: new[] { "PartitionKey", "RowKey" });
int deleted = 0;
await foreach (var entity in query)
{
try
{
await tableClient.DeleteEntityAsync(entity.PartitionKey, entity.RowKey);
deleted++;
}
catch (Exception ex)
{
_logger.LogWarning(ex, "Failed to delete audit log {PK}/{RK}", entity.PartitionKey, entity.RowKey);
}
}
var deleted = 0;
deleted += await DeleteAuditLogsByFilterAsync(tableClient, BusinessTimestamp.AuditRetentionClause(cutoffUtc));
// 'RowKey ge '0'' cannot match any '!'-row (0x21 < 0x30), only legacy GUIDs.
deleted += await DeleteAuditLogsByFilterAsync(tableClient,
$"RowKey ge '0' and Timestamp lt datetime'{cutoffUtc:yyyy-MM-ddTHH:mm:ss}Z'");
if (deleted > 0)
_logger.LogInformation("Deleted {Count} audit log entries older than {Cutoff:yyyy-MM-dd}", deleted, cutoffUtc);
@@ -110,6 +108,25 @@ namespace AutopilotMonitor.Functions.Services
}
}
private async Task<int> DeleteAuditLogsByFilterAsync(TableClient tableClient, string filter)
{
var query = tableClient.QueryAsync<TableEntity>(filter: filter, select: new[] { "PartitionKey", "RowKey" });
int deleted = 0;
await foreach (var entity in query)
{
try
{
await tableClient.DeleteEntityAsync(entity.PartitionKey, entity.RowKey);
deleted++;
}
catch (Exception ex)
{
_logger.LogWarning(ex, "Failed to delete audit log {PK}/{RK}", entity.PartitionKey, entity.RowKey);
}
}
return deleted;
}
/// <summary>
/// Gets audit log entries for a tenant within an optional UTC date window.
/// No row cap — returns the full filtered set, sorted newest-first.
@@ -384,9 +401,9 @@ namespace AutopilotMonitor.Functions.Services
if (!string.IsNullOrEmpty(lastRowKey))
clauses.Add($"RowKey gt '{lastRowKey!.Replace("'", "''")}'");
if (dateFrom.HasValue)
clauses.Add($"Timestamp ge datetime'{ToUtc(dateFrom.Value):o}'");
clauses.Add(BusinessTimestamp.AuditDateFromClause(ToUtc(dateFrom.Value)));
if (dateTo.HasValue)
clauses.Add($"Timestamp le datetime'{ToUtc(dateTo.Value):o}'");
clauses.Add(BusinessTimestamp.AuditDateToClause(ToUtc(dateTo.Value)));
return string.Join(" and ", clauses);
}
@@ -398,14 +415,30 @@ namespace AutopilotMonitor.Functions.Services
EntityType = entity.GetString("EntityType") ?? string.Empty,
EntityId = entity.GetString("EntityId") ?? string.Empty,
PerformedBy = entity.GetString("PerformedBy") ?? string.Empty,
Timestamp = entity.GetDateTimeOffset("Timestamp")?.UtcDateTime ?? DateTime.UtcNow,
Timestamp = ResolveAuditTimestamp(entity),
Details = entity.GetString("Details") ?? string.Empty,
};
// The system Timestamp is auto-managed by Azure on insert and reliably
// sortable; the user-defined "Timestamp" property is set by LogAuditEntry
// for parity with the model. We filter on the system Timestamp since it
// is always indexed.
/// <summary>
/// Business time of an audit row: OccurredUtc column → time decoded from the
/// reverse-tick RowKey → system Timestamp. The system Timestamp is only a last
/// resort (legacy bare-GUID rows): it is reset by any row rewrite, so for rows
/// migrated between storage accounts it holds the migration moment, not the
/// event time.
/// </summary>
internal static DateTime ResolveAuditTimestamp(TableEntity entity)
{
var occurred = BusinessTimestamp.GetUtcDateTime(entity, BusinessTimestamp.OccurredUtcColumn);
if (occurred.HasValue)
return occurred.Value;
if (BusinessTimestamp.TryDecodeAuditRowKey(entity.RowKey, out var decoded))
return decoded;
return BusinessTimestamp.GetUtcDateTime(entity, "Timestamp") ?? DateTime.UtcNow;
}
// Date windows filter on the RowKey (reverse-tick, index-backed) rather than any
// timestamp property: the system Timestamp is reset by migrations, and an
// OccurredUtc property filter would exclude rows that lack the column.
internal static string? BuildAuditLogFilter(string? tenantId, DateTime? dateFrom, DateTime? dateTo,
bool excludeDeletions = false, AuditLogQueryFilters? filters = null)
{
@@ -424,11 +457,11 @@ namespace AutopilotMonitor.Functions.Services
}
if (dateFrom.HasValue)
{
clauses.Add($"Timestamp ge datetime'{ToUtc(dateFrom.Value):o}'");
clauses.Add(BusinessTimestamp.AuditDateFromClause(ToUtc(dateFrom.Value)));
}
if (dateTo.HasValue)
{
clauses.Add($"Timestamp le datetime'{ToUtc(dateTo.Value):o}'");
clauses.Add(BusinessTimestamp.AuditDateToClause(ToUtc(dateTo.Value)));
}
return string.Join(" and ", clauses);
}
@@ -681,7 +681,10 @@ namespace AutopilotMonitor.Functions.Services
["EventId"] = evt.EventId,
["SessionId"] = evt.SessionId,
["TenantId"] = evt.TenantId,
["Timestamp"] = evt.Timestamp,
// "Timestamp" is a reserved system property — supplied values are ignored
// and any row rewrite (storage migration) resets it. The sanitized agent
// event time therefore lives in its own column.
[BusinessTimestamp.OccurredUtcColumn] = EnsureUtc(evt.Timestamp),
["EventType"] = evt.EventType ?? string.Empty,
["Severity"] = (int)evt.Severity,
["Source"] = evt.Source ?? string.Empty,
@@ -777,7 +780,8 @@ namespace AutopilotMonitor.Functions.Services
["EventId"] = evt.EventId,
["SessionId"] = evt.SessionId,
["TenantId"] = evt.TenantId,
["Timestamp"] = evt.Timestamp,
// Reserved-name caveat: see StoreEventAsync.
[BusinessTimestamp.OccurredUtcColumn] = EnsureUtc(evt.Timestamp),
["EventType"] = evt.EventType ?? string.Empty,
["Severity"] = (int)evt.Severity,
["Source"] = evt.Source ?? string.Empty,
@@ -2673,10 +2677,7 @@ namespace AutopilotMonitor.Functions.Services
EventId = entity.GetString("EventId") ?? string.Empty,
SessionId = entity.GetString("SessionId") ?? string.Empty,
TenantId = tenantId ?? partitionKey,
Timestamp = DateTime.SpecifyKind(
entity.GetDateTimeOffset("Timestamp")?.UtcDateTime
?? entity.GetDateTime("Timestamp")
?? DateTime.UtcNow, DateTimeKind.Utc),
Timestamp = DateTime.SpecifyKind(ResolveEventTimestamp(entity), DateTimeKind.Utc),
EventType = entity.GetString("EventType") ?? string.Empty,
Severity = (EventSeverity)(entity.GetInt32("Severity") ?? 0),
Source = entity.GetString("Source") ?? string.Empty,
@@ -2827,7 +2828,10 @@ namespace AutopilotMonitor.Functions.Services
/// <summary>
/// Returns the earliest event timestamp persisted for a session, if any.
/// Events are written with RowKey "{Timestamp}_{Sequence}", so querying the partition
/// and taking the first row yields the earliest event.
/// and taking the first row yields the earliest event. Resolution must NOT rely on
/// the system Timestamp: it is reset by storage migrations, and this value feeds the
/// StartedAt/Duration reconciliation — a migration-reset read would rewrite those
/// session fields with the migration date.
/// </summary>
private async Task<DateTime?> GetEarliestSessionEventTimestampAsync(string tenantId, string sessionId)
{
@@ -2838,13 +2842,12 @@ namespace AutopilotMonitor.Functions.Services
var query = tableClient.QueryAsync<TableEntity>(
filter: $"PartitionKey eq '{partitionKey}'",
maxPerPage: 1,
select: new[] { "Timestamp", "RowKey" }
select: new[] { "RowKey", BusinessTimestamp.OccurredUtcColumn, "Timestamp" }
);
await foreach (var entity in query)
{
return entity.GetDateTimeOffset("Timestamp")?.UtcDateTime
?? entity.GetDateTime("Timestamp");
return ResolveEventTimestampOrNull(entity);
}
}
catch (Exception ex)
@@ -2855,6 +2858,26 @@ namespace AutopilotMonitor.Functions.Services
return null;
}
/// <summary>
/// Business time of an event row: OccurredUtc column → agent time decoded from the
/// "{yyyyMMddHHmmssfff}_{seq}" RowKey → system Timestamp (last resort — reset by any
/// row rewrite, e.g. a storage migration; for migrated rows it holds the migration
/// moment). Event rows are deliberately never backfilled, so the RowKey decode is the
/// primary path for all pre-OccurredUtc rows.
/// </summary>
internal static DateTime ResolveEventTimestamp(TableEntity entity)
=> ResolveEventTimestampOrNull(entity) ?? DateTime.UtcNow;
internal static DateTime? ResolveEventTimestampOrNull(TableEntity entity)
{
var occurred = BusinessTimestamp.GetUtcDateTime(entity, BusinessTimestamp.OccurredUtcColumn);
if (occurred.HasValue)
return occurred.Value;
if (BusinessTimestamp.TryDecodeEventRowKeyPrefix(entity.RowKey, out var decoded))
return decoded;
return BusinessTimestamp.GetUtcDateTime(entity, "Timestamp");
}
/// <summary>
/// Returns the authoritative (total, reboot) event-row counts for a session, or <c>null</c>
/// if they could not be determined (storage error). Events are written with UpsertReplace
@@ -1,7 +1,6 @@
"use client";
import { Suspense, useCallback, useEffect, useMemo, useRef, useState } from "react";
import Link from "next/link";
import { useRouter, useSearchParams } from "next/navigation";
import { ProtectedRoute } from "../../components/ProtectedRoute";
import { GlobalAdminBanner } from "@/components/GlobalAdminBanner";
@@ -31,6 +31,9 @@ export function SignalRProvider({ children }: { children: React.ReactNode }) {
const [joinedGroups, setJoinedGroups] = useState<string[]>([]);
const retryCountRef = useRef(0);
const retryTimeoutRef = useRef<ReturnType<typeof setTimeout> | null>(null);
// Collapses rapid successive accessRevoked pushes (one per revoked grant) into a single
// restart — a second start() on a non-Disconnected connection throws and burns the retry budget.
const revokeRestartInFlightRef = useRef(false);
// Set in onreconnecting, read in onreconnected so we can report downtime as a measurement.
const disconnectStartedAtRef = useRef<number | null>(null);
const maxRetries = 3;
@@ -157,12 +160,20 @@ export function SignalRProvider({ children }: { children: React.ReactNode }) {
// the next Connected transition — through the join endpoint, which re-runs authorization
// against the fresh scope. Revoked groups 403; still-authorized streams recover automatically.
newConnection.on('accessRevoked', async () => {
// Dropping an event that arrives mid-restart is safe: the rejoin after start() re-runs
// authorization against the CURRENT scope, which already reflects that later revoke.
if (revokeRestartInFlightRef.current) return;
revokeRestartInFlightRef.current = true;
console.warn('[SignalR] accessRevoked received — restarting connection to re-authorize group memberships');
trackEvent('signalr_access_revoked');
try {
await newConnection.stop();
} catch { /* proceed to restart regardless */ }
startConnection();
try {
await startConnection();
} finally {
revokeRestartInFlightRef.current = false;
}
});
// Start as soon as the user is authenticated. The previous 2s setTimeout