Skip to content

armgit / Pulse

Public Imported from GitHub
0

Real-time monitoring dashboard for Proxmox VE, PBS, Docker, Kubernetes, TrueNAS and vSphere. Self-hosted, with alerts for failed backups, full disks, unhealthy containers and offline nodes.

Explain saved infrastructure access recovery without deleting state

pulse-triage[bot]77edfedcb27f
diff --git a/docs/TROUBLESHOOTING.md b/docs/TROUBLESHOOTING.md
index 8c3ae55e81..d508347c90 100644
--- a/docs/TROUBLESHOOTING.md
+++ b/docs/TROUBLESHOOTING.md
@@ -21,6 +21,10 @@ establish safe recovery.
or [401 and login failures](TROUBLESHOOTING.md#cannot-login--401-unauthorized),
according to the existing authentication method. Do not delete configuration
or repeat first-time setup.
+- **Signed in, but saved infrastructure settings are unavailable:** use
+ [saved-settings access recovery](#signed-in-but-saved-infrastructure-settings-are-unavailable).
+ An unreadable connection is not evidence that it was removed; keep its saved
+ credentials, TLS trust and collection scope.
- **Server installation finished but Pulse is unavailable:** distinguish
[container state, Pulse service state and web access](INSTALL.md#installation-did-not-reach-a-usable-server).
Preserve the existing installation; a success banner is not readiness.
@@ -518,6 +522,48 @@ full headers, **Copy as cURL** commands and network exports private. A 401 alone
does not justify deleting configuration, recreating the data volume or
re-enrolling agents.
+#### Signed in, but saved infrastructure settings are unavailable
+
+In **Settings → Infrastructure → Manage**, an unavailable saved configuration
+means Pulse could not read the matching connection for the current account and
+organisation. It does **not** establish that the connection was removed, that
+its saved password is lost, or that the monitored server rejected it. Monitoring
+access and permission to manage saved connections are different.
+
+1. Check that you are using the intended Pulse URL, account and organisation.
+ Preserve any working administrator session. After an account, organisation
+ or access change, Pulse withdraws stale fields and test results rather than
+ reuse another reader's settings. Do not copy values back from an old tab.
+2. Use **Retry saved configuration**, when shown, to read the saved settings
+ again before editing. This is a read, not a save, connection test or removal.
+ If the page says **Reload before making changes**, restore the intended
+ access first, then reload that page; do not repeat a refused Test or Save.
+3. If access is still refused, ask the Pulse administrator to check the
+ account's intended role, organisation and connection access. A successful
+ sign-in or visible dashboard does not grant settings access. Keep existing
+ restrictions: do not widen token scopes, grant another department's nodes,
+ disable authentication or bypass CSRF checks just to make the error go away.
+
+An HTTP **401** concerns the Pulse session or token; follow the
+[login guidance](#cannot-login--401-unauthorized). An HTTP **403** can be an
+access refusal or a rejected browser write's CSRF check; identify the request
+and whether it was a read or a write, not just the status. A failed read does
+not prove an empty inventory. A timeout or lost Save response can occur after
+a change was applied: check the existing state before any further write.
+
+Do not remove and recreate the connection, reinstall an agent or enter
+replacement secrets to recover an unread form. Preserve saved passwords/API
+keys, certificate verification, fingerprints and collection scope. A successful
+read restores the editor, not proof of polling or recovery of missing readings;
+check those separately using the [affected workflow](#choose-the-affected-workflow).
+
+For a report, keep the time, Pulse version, platform, action, redacted error and
+request path without query strings, plus whether the failure followed an
+account or organisation change. Use observations already available; do not
+run Diagnostics or replay a mutation just to collect them. Keep tokens,
+cookies, passwords, full configuration, **Copy as cURL** and network exports
+private; a cropped error must not expose saved credentials or private hosts.
+
#### Audit Log verification shows unsigned events
**Unsigned** means no signature was stored for that event; it is not the same
diff --git a/frontend-modern/public/docs/TROUBLESHOOTING.md b/frontend-modern/public/docs/TROUBLESHOOTING.md
index 8c3ae55e81..d508347c90 100644
--- a/frontend-modern/public/docs/TROUBLESHOOTING.md
+++ b/frontend-modern/public/docs/TROUBLESHOOTING.md
@@ -21,6 +21,10 @@ establish safe recovery.
or [401 and login failures](TROUBLESHOOTING.md#cannot-login--401-unauthorized),
according to the existing authentication method. Do not delete configuration
or repeat first-time setup.
+- **Signed in, but saved infrastructure settings are unavailable:** use
+ [saved-settings access recovery](#signed-in-but-saved-infrastructure-settings-are-unavailable).
+ An unreadable connection is not evidence that it was removed; keep its saved
+ credentials, TLS trust and collection scope.
- **Server installation finished but Pulse is unavailable:** distinguish
[container state, Pulse service state and web access](INSTALL.md#installation-did-not-reach-a-usable-server).
Preserve the existing installation; a success banner is not readiness.
@@ -518,6 +522,48 @@ full headers, **Copy as cURL** commands and network exports private. A 401 alone
does not justify deleting configuration, recreating the data volume or
re-enrolling agents.
+#### Signed in, but saved infrastructure settings are unavailable
+
+In **Settings → Infrastructure → Manage**, an unavailable saved configuration
+means Pulse could not read the matching connection for the current account and
+organisation. It does **not** establish that the connection was removed, that
+its saved password is lost, or that the monitored server rejected it. Monitoring
+access and permission to manage saved connections are different.
+
+1. Check that you are using the intended Pulse URL, account and organisation.
+ Preserve any working administrator session. After an account, organisation
+ or access change, Pulse withdraws stale fields and test results rather than
+ reuse another reader's settings. Do not copy values back from an old tab.
+2. Use **Retry saved configuration**, when shown, to read the saved settings
+ again before editing. This is a read, not a save, connection test or removal.
+ If the page says **Reload before making changes**, restore the intended
+ access first, then reload that page; do not repeat a refused Test or Save.
+3. If access is still refused, ask the Pulse administrator to check the
+ account's intended role, organisation and connection access. A successful
+ sign-in or visible dashboard does not grant settings access. Keep existing
+ restrictions: do not widen token scopes, grant another department's nodes,
+ disable authentication or bypass CSRF checks just to make the error go away.
+
+An HTTP **401** concerns the Pulse session or token; follow the
+[login guidance](#cannot-login--401-unauthorized). An HTTP **403** can be an
+access refusal or a rejected browser write's CSRF check; identify the request
+and whether it was a read or a write, not just the status. A failed read does
+not prove an empty inventory. A timeout or lost Save response can occur after
+a change was applied: check the existing state before any further write.
+
+Do not remove and recreate the connection, reinstall an agent or enter
+replacement secrets to recover an unread form. Preserve saved passwords/API
+keys, certificate verification, fingerprints and collection scope. A successful
+read restores the editor, not proof of polling or recovery of missing readings;
+check those separately using the [affected workflow](#choose-the-affected-workflow).
+
+For a report, keep the time, Pulse version, platform, action, redacted error and
+request path without query strings, plus whether the failure followed an
+account or organisation change. Use observations already available; do not
+run Diagnostics or replay a mutation just to collect them. Keep tokens,
+cookies, passwords, full configuration, **Copy as cURL** and network exports
+private; a cropped error must not expose saved credentials or private hosts.
+
#### Audit Log verification shows unsigned events
**Unsigned** means no signature was stored for that event; it is not the same
diff --git a/scripts/tests/test_login_troubleshooting_docs.py b/scripts/tests/test_login_troubleshooting_docs.py
index b9d28c4fc4..6d16bfbadd 100644
--- a/scripts/tests/test_login_troubleshooting_docs.py
+++ b/scripts/tests/test_login_troubleshooting_docs.py
@@ -18,6 +18,43 @@ def section(heading: str) -> str:
class LoginTroubleshootingDocsTest(unittest.TestCase):
+ def test_unread_saved_settings_are_not_treated_as_removed_connections(self):
+ text = section('#### Signed in, but saved infrastructure settings are unavailable')
+ for expected in (
+ 'Settings → Infrastructure → Manage', 'current account and organisation',
+ 'does **not** establish that the connection was removed',
+ 'Monitoring access and permission to manage saved connections are different',
+ 'Retry saved configuration', 'read, not a save, connection test or removal',
+ 'Reload before making changes', 'do not repeat a refused Test or Save',
+ 'do not widen token scopes', 'bypass CSRF checks',
+ 'A failed read does not prove an empty inventory',
+ 'check the existing state before any further write',
+ 'Do not remove and recreate the connection',
+ 'replacement secrets', 'certificate verification, fingerprints and collection scope',
+ 'not proof of polling', 'request path without query strings',
+ 'do not run Diagnostics or replay a mutation', 'network exports private',
+ ):
+ with self.subTest(statement=expected):
+ self.assertIn(expected, text)
+ doc = (ROOT / 'docs/TROUBLESHOOTING.md').read_text(encoding='utf-8')
+ quick = doc.split('## Choose the affected workflow\n', 1)[1].split('\n## ', 1)[0]
+ self.assertIn('#signed-in-but-saved-infrastructure-settings-are-unavailable', quick)
+
+ def test_saved_settings_recovery_matches_the_read_only_ui_action(self):
+ workspace = (ROOT / 'frontend-modern/src/components/Settings/InfrastructureWorkspace.tsx').read_text()
+ panel = workspace.split('const renderUnavailableSavedConfiguration =', 1)[1].split(
+ 'const renderConnectionSlot =', 1)[0]
+ self.assertIn('This does not confirm removal.', panel)
+ self.assertIn('Retry saved configuration', panel)
+ self.assertIn('onClick={() => void state.loadConnections()}', panel)
+ for action in ('saveConnection(', 'testConnection(', 'deleteConnection('):
+ self.assertNotIn(action, panel)
+ for platform in ('VMware', 'TrueNAS'):
+ hook = (ROOT / f'frontend-modern/src/components/Settings/use{platform}SettingsPanelState.ts').read_text()
+ self.assertIn('if (![401, 403].includes(apiErrorStatus(error) ?? 0)) return false;', hook)
+ self.assertIn('Reload before making changes.', hook)
+ self.assertIn("eventBus.on('org_switched'", hook)
+
def test_local_credentials_use_actual_precedence_and_file_format(self):
text = section('#### "Invalid username or password" after setup')
for expected in (