DNS External Resolution TSG: quick-fix, full admin-surface discoverability, validator table - #332
Conversation
…ility, validator table - Add a 'Quick fix (start here)' block: auto-identify the management adapter and give the exact reversible Set-DnsClientServerAddress command, gated by 'do not guess DNS servers', so the ~90% case is unblocked in one screen without reading the full decision tree first. - Add a 'Where it appears across the admin surfaces' subsection covering all eight admin surfaces (node PowerShell, event log, portal, on-disk result JSON = shown; Get-ClusterLog, Failover Cluster Manager, WAC standalone, WAC-in-portal = explicitly not evident), so admins do not hunt in the wrong place. - Add a validator metadata table (Name / Validator-test / Component / Severity) under the H1. No change to the diagnosis or remediation logic; documentation completeness and time-to-fix only.
There was a problem hiding this comment.
Pull request overview
Expands the external DNS resolution TSG to improve remediation speed and failure discoverability.
Changes:
- Adds validator metadata.
- Introduces a quick-fix PowerShell workflow.
- Documents visibility across eight administrative surfaces.
…elect, per-server verify, template table - Quick fix now distinguishes deployment/add-node (node DNS-client change OK) from an already-deployed cluster, where changing DNS servers is UNSUPPORTED post-deployment; deployed clusters are directed to the supported upstream DNS/forwarder fix. - Management adapter is selected by the node's known management IP and fails closed unless exactly one up adapter owns it (multihomed nodes: no enumeration-order guessing). - Verify step now queries every configured server with -Server, matching the validator, so a working default resolver cannot mask a still-failing server. - Metadata block converted to the Troubleshoot-Template HTML table with Applicable Scenarios and Affected Versions.
|
I checked the new content against the product source and the Azure Local docs, and ran the lint. On the facts it holds up. The identifiers are all real: One consistency issue is worth fixing before merge. The new Quick fix correctly tells a deployed-cluster operator not to re-point the node's DNS client (unsupported post-deployment) and sends them to the upstream forwarder fix. The existing Remediation section still says the opposite: "Most common fix (start here)" routes to step 3 first option, which re-points the DNS client with no deployment caveat, and the closing line calls re-pointing a [LOW RISK] change with no qualification. The Overview says the deployed pre-update case is the one you will most often hit, so the most common reader is being pointed at the action the Quick fix says is unsupported and can break name resolution for the whole node, and it is labeled [LOW RISK]. Scope that pre-existing re-point option, the [LOW RISK] closing line, and the "At a glance" DNS-client-fix framing to deployment and add-node only, matching the Quick fix, and lead the deployed case to the forwarder option. [HIGH RISK] (The lint flags a PowerShell parse error on the |
…self-contradiction) The Quick fix section correctly states that on an already-deployed cluster the node DNS client must not be changed (unsupported post-deployment) and routes the fix upstream to a forwarder. The Remediation section still framed re-pointing the node DNS client as the [LOW RISK] 'Most common fix (start here)' with no such caveat, so a deployed-cluster operator reading Remediation was routed to the forbidden action. Reconciled all three spots (intro paragraph, step 3 first option, and the risk-label closing) to mirror the Quick fix support boundary: re-point node DNS only at deployment / add-node time; on a deployed cluster use the upstream DNS server / forwarder change.
Summary
Documentation-completeness improvements to
TSG/EnvironmentValidator/Troubleshooting-DNS-External-DNS-Resolution.md. No change to the diagnosis or remediation logic — this makes the common fix faster to reach and rounds out where the failure is (and is not) visible.Changes
Set-DnsClientServerAddresscommand for the ~90% case (a node pointed at a DNS server that cannot resolve external names), gated by a bold "do not guess DNS server IP addresses" so a literal follower stays safe. Readers who need the full decision tree are pointed to Remediation.HealthCheckResultJSON are shown;Get-ClusterLog, Failover Cluster Manager, and Windows Admin Center (standalone and in-portal) are called out as not evident, so admins do not hunt in the wrong tool.Invoke-AzStackHciDNSValidation -Include Test-ExternalDnsResolution.Why
Testing