Backport 6375d09a8b. Candidate f64fbb8b7 reproduced six false independent confirmations before repair; focused Docker result/executor and receipt reconciliation tests pass afterwards. Preserve stopped updates and execution history. This does not resolve issue #1891 or qualify an installed artifact.
Change-source: pulse-maintainer
Backport production fixes 57cb1a83bf and 49b998a251, with restart coverage from 49928385ba. Adapt PBS webhook assertions to the release-line nonzero recovery fixture.
On candidate 6558b2d106, reproduced confirmed-empty storage retaining its incident; applying only the storage guard then reproduced stale recovery content. Both fixes pass focused alert checks and three race-enabled repetitions across alerts and monitoring. Local synthetic receiver and orderly restart evidence only; no installed qualification.
Change-source: pulse-maintainer
Backport of 31f1f1933a, adapted to release/v6.4 client tests. Reproduced absent envelopes returning zero metrics and the null-status poller availability failure on candidate base 239ee0f9da before applying the production fix. Focused client and lifecycle race checks pass three repetitions. No installed qualification claimed.
Change-source: pulse-maintainer
Callback-only PBS lifecycle coverage cannot detect broken monitor wiring or notification queue delivery. Exercise the real poller and production callbacks through a local receiver, requiring a successful firing audit before recovery and retaining incident identity during node endpoint failures.
Change-source: pulse-maintainer
(cherry picked from commit a815f8bb02)
Exercise real threshold incidents with SQLite active state across two manager restarts. Missing metrics retain identity without recovery; measured zero resolves each incident once and durable history survives. Disabling the availability guard makes the regression fail.
Change-source: pulse-maintainer
(cherry picked from commit fe563325c2)
Backport 2b78867cae with the PBS HTTP fixture and connectivity characterisation from a01efeed6b. Exclude unrelated main-line security-status documentation.
Release-line reproduction on 369e7f339e: node-status denial falsely resolves active memory utilisation. Retain active metric alerts until a valid sample arrives; connectivity and policy suppression remain independent.
Validation: go test -race ./internal/alerts ./internal/models ./internal/monitoring -run 'Test.*PBS' -count=3 passes. Installed destination delivery and restart persistence remain unqualified.
Change-source: pulse-maintainer
Reproduced on release base 11539a8059: HTTP 502 quoting API error 403 and HTTP 503 quoting API error 404 discarded cached backups. Both now preserve inventory; genuine 401/403/404 remain terminal.
Backport of c3b28f4557, including substantive monitoring contract and verification documentation. Adapt client test rename to the three existing callers; no unrelated mainline tests imported.
Validation: focused monitoring and PBS client tests passed with -race -count=3. No installed candidate, notification delivery or soak claim.
Change-source: pulse-maintainer
Protect the existing #1899 release-line regression backport with independent hydration and resource-snapshot signals. Both boundary cases reject a mutation to general hydration; 88 focused tests pass. No runtime change.
Change-source: pulse-maintainer
Incorporate GitHub's merge of the prior reviewed CPU routing guard while
preserving the exact unequal-window regression backport and its ancestry.
Change-source: pulse-maintainer
Equal-cadence CPU regression checks can miss a shared cached percentage because both reporting loops expect the same average. Add unequal host and Docker intervals and replace the Docker collector to guard independent baseline ownership without resetting the host loop. This protects the existing issue #1894 repair; synthetic coverage does not establish installed-agent qualification or reporter resolution.
Contract-Neutral: regression-test coverage only; no runtime behaviour or product surface changes.
Change-source: pulse-maintainer
(cherry picked from commit fc80229b1b)
Issue #1894 exposed CPU sampling interference between host and Docker reporting loops. Existing tests cover independent collectors and host routing but do not exercise Docker's production dependency entry points.
Interleave the real host, Docker and package-level collectors over synthetic procfs counters. Alternate Docker's include and no-include paths and require each loop to retain a full-interval 5 percent baseline. Production behaviour is unchanged.
Validation: focused test passed; reverting either Docker route to the package-level collector made it fail. After restoration, three focused CPU tests passed with -race -count=3 across dockeragent, hostagent and hostmetrics. git diff --check passed.
Change-source: pulse-maintainer
Backport-of: c590a01f33
Protect the existing release-line repair d33f81709e for candidate regression #1882 across repeated registry rebuilds and SQLite close/reopen. No runtime or product scope change.
Validation: focused test passes; restoring rebuild timestamps fails with 18 samples instead of 3; lifecycle and canonical-storage tests pass three race repetitions.
Contract-Neutral: regression coverage for existing observation-time behaviour.
Change-source: pulse-maintainer
Preserve the reviewed reconnect navigation backport while incorporating the canonical merge of the critical-transition alert repair.
Change-source: pulse-maintainer
Reproduced on release candidate 7f44ccd047: WARNING to EMERGENCY retained severity but dispatched only once instead of twice. Backport the existing delivery repair and acknowledgement/rate-budget lifecycle regression coverage; no new product scope.
Backport-of: 4752d5b4b7 (code and tests), b116d05592
Validation: five focused race repetitions passed; pre-fix active transition failed. Appliance/provider receipt and restarted candidate soak remain outstanding.
Change-source: pulse-maintainer
Backport runtime mapping and regression tests from 65a024b3d3. Candidate v6.4.3-rc.1 and stable v6.4.1 omit the documented EMERGENCY severity. Include release-line monitoring and unified-resource contracts for the repaired projection and recovery boundary.
Reproduced omitted projection, zero notification callbacks and false recovery before the fix. Affected package suites and focused race tests passed in the initial turn. Appliance and external-provider receipt remain unqualified.
Change-source: pulse-maintainer
The INFO and NOTICE repairs need lifecycle coverage beyond isolated severity transitions. Verify that escalation retains identity and that renewed actionable evidence interrupts recovery, preventing a later single INFO observation from clearing the incident.
Change-source: pulse-maintainer
(cherry picked from commit 62c66d2a48)
Correct the INFO backport so native TrueNAS NOTICE remains actionable while retaining informational canonical severity. Preserve the provider level as optional evidence and pin projection, JSON compatibility, alert activation, and confirmed recovery boundaries.
Adapted from ff1d010dd9 for release/v6.4.
Change-source: pulse-maintainer
Issue #1892 reports successful replication information requiring acknowledgement. Preserve provider information on resources while excluding TrueNAS INFO-level conditions from active alert synchronisation. Warning conditions still activate and clear when downgraded to information.
Resource-incident timeline events contain numeric placeholders, not threshold evidence. Avoid displaying these as a fictitious 0 >= 0 trigger while retaining numeric metric formatting. Focused regressions reproduce both failures and pass with these changes.
Change-source: pulse-maintainer
(cherry picked from commit 7726539f83)
TrueNAS emits FINISHED for successful replication, but the recovery mapper treated it as unknown. Recognise that provider state while preserving error precedence and the missing-run guard; add regression coverage reproducing the reported outcome.
Change-source: pulse-maintainer
(cherry picked from commit 62c6e07477)
Issue #1895 reports parity alerts when mdNumDisks=0 on a pool-only Unraid system. Array service state alone does not establish that a parity array exists.
Preserve the optional disk count from collection through canonical runtime conversion and suppress only the no-parity warning for an explicit zero. Missing or malformed counts retain legacy behaviour, and disk failure reasons remain active.
Validated focused Unraid tests in hostagent, storagehealth, monitoring, unifiedresources and alerts, including JSON zero preservation and canonical round trip. The new pool-only regression fails against the previous warning condition. Both agent and server need this change; no release or reporter retest is claimed.
Change-source: pulse-maintainer
(cherry picked from commit 3334cccfd9)
Near-synchronous host and Docker reports consumed a shared CPU baseline, measuring collection bursts rather than each module's reporting interval. Retain a collector per host collector and a separate Docker module collector while preserving the package-level convenience API and disk filters.
Add an interleaved-counter regression covering both collection entry points. It fails when routed through the shared baseline and passes with isolated state. All hostmetrics, hostagent and dockeragent tests pass, as do the focused CPU regression tests under the race detector.
Change-source: pulse-maintainer
(cherry picked from commit 0f972f42f2)
Issue #1890 reports macOS agent updates stopping because the root group does not exist. Use numeric superuser ownership in the two shared lifecycle writes without relaxing failure handling or the least-privilege group boundary. Add a regression fixture that rejects named root ownership and checks that chown failures still prevent replacement.
Change-source: pulse-maintainer
(cherry picked from commit 19c2b6a925)
The release line carries an additional oversized-metadata regression fixture. Model its published archive HEAD response so the backported existence check retains coverage without treating that fixture as a deleted release.
Refs: #1881
Contract-Neutral: backport test adaptation only
Change-source: pulse-maintainer
GitHub's Atom feed can retain entries for deleted releases. Probe the deterministic runtime archive without following redirects before exposing a feed candidate, and continue to the next published version when the candidate is stale. Keep custom update servers from falling through to the public feed on a 403.
Contract-Neutral: Update discovery validation only; no wire, persisted-data, or subsystem contract change
Refs: #1881
Change-source: pulse-maintainer
(cherry picked from commit 3193db66b0)
GitHub's release list now exceeds the update check's 1 MiB safety limit,
causing valid installations to report that updates are unavailable. Keep the
bound and use the existing bounded Atom feed only for the typed over-limit
condition so malformed metadata remains a hard error.
Accept both release-name and bare-tag feed entries, and do not substitute the
public GitHub feed for a configured custom update server.
Contract-Neutral: Update discovery error handling only; no wire, persisted-data, or subsystem contract change
Refs: #1881
Change-source: pulse-maintainer
(cherry picked from commit a186bd205e)
Operators can explicitly select tmpfs paths that automatic filesystem
filtering would otherwise hide. Multiple such mounts commonly share both the
generic tmpfs device name and capacity, so applying normal storage
deduplication to them silently dropped configured disks from collection.
Limit device-and-capacity deduplication to normally visible storage while
preserving it for NAS bind mounts and subvolumes.
Change-source: issue #1875
Backport of the main-branch guard. pkg/server tests boot the real server
through Run() with the version literal "test-version", which normalizes to
0.0.0-test-version, and each test runs against its own t.TempDir(), so
every run mints a fresh install ID. The service-health failure reporter
sends synchronously from a deferred handler as soon as Run() returns an
error, so any test exercising a startup failure posts one ping.
This line still emitted after main was fixed: release-line lane work runs
pkg/server tests on this branch, and those pings arrive with the old
version classifier too, so they land mislabelled as ordinary prereleases
and re-contaminate install-population reads that were just corrected.
send() now refuses the production endpoint whenever testing.Testing()
reports true. The check compares against productionPingEndpoint, so
telemetry's own tests keep asserting on ping content through a redirected
endpoint. Verified on this branch: three runs of the failing-startup
tests, zero pings received.
Backport of the main-branch guard. pkg/server tests boot the real server
through Run() with the version literal "test-version", which normalizes to
0.0.0-test-version, and each test runs against its own t.TempDir(), so
every run mints a fresh install ID. The service-health failure reporter
sends synchronously from a deferred handler as soon as Run() returns an
error, so any test exercising a startup failure posts one ping.
This line still emitted after main was fixed: release-line lane work runs
pkg/server tests on this branch, and those pings arrive with the old
version classifier too, so they land mislabelled as ordinary prereleases
and re-contaminate install-population reads that were just corrected.
send() now refuses the production endpoint whenever testing.Testing()
reports true. The check compares against productionPingEndpoint, so
telemetry's own tests keep asserting on ping content through a redirected
endpoint. Verified on this branch: three runs of the failing-startup
tests, zero pings received.
Canonical registry rebuilds can run several times between PBS polls. Reuse the metric source sighting time so unchanged datastore capacity does not create a fresh raw sample on every rebuild.
Contract-Neutral: behavioral regression fix; metric contract and agent lifecycle boundaries unchanged
Change-source: pulse-maintainer
Exclude the SSH connection deadlock advisories published against x/crypto versions before v0.56.0.
Contract-Neutral: dependency-only security floor update; no runtime contract changed
Change-source: pulse-maintainer
GitHub's releases API response now exceeds the updater's 1 MiB
metadata limit. This causes v6.4.3-rc.1 update checks to fail even
though the releases Atom feed remains bounded.
Retain the safety cap and use the existing bounded Atom path only for
the default GitHub source. Custom update servers remain fail-closed.
Contract-Neutral: update-source recovery fix only; no public contract change
Change-source: pulse-maintainer
pulse-triage[bot]