Fix for delete lb and stale lb dsr vfp rules.

Author: princepereira

scripts/mitigation-scripts/stale-lb-dsr-rules/README.md

@@ -0,0 +1,132 @@
+# Stale LB DSR Rules Cleanup
+
+## Overview
+
+This mitigation script automatically detects and removes stale Load Balancer Direct Server Return (LB DSR) rules from VFP (Virtual Filtering Platform) that reference non-existent backend endpoints. It runs continuously to maintain network health by cleaning up orphaned rules that can cause connectivity issues.
+
+## Problem Statement
+
+When backend endpoints are removed or become unavailable, the corresponding LB DSR rules in VFP may not be cleaned up properly. These stale rules can:
+- Cause packet routing failures
+- Lead to connection timeouts
+- Create unnecessary overhead in the networking stack
+- Result in traffic being sent to non-existent endpoints
+
+## Solution
+
+The `cleanup-stale-lb-rules.ps1` script:
+1. Checks and sets the required registry configuration for LB DSR feature management
+2. Continuously monitors VFP LB DSR rules (both IPv4 and IPv6)
+3. Compares rule destination IPs (DIPs) against active HNS endpoints
+4. Automatically removes rules that reference non-existent endpoints
+
+## Prerequisites
+
+- Windows Server with HNS (Host Network Service) enabled
+- VFP control utilities (`vfpctrl.exe`) available
+- PowerShell with administrator privileges
+- HNS PowerShell module
+
+## Usage
+
+### Running the Script on a Single Node
+
+```powershell
+.\cleanup-stale-lb-rules.ps1
+```
+
+The script will:
+1. Check registry key `HKLM:\SYSTEM\CurrentControlSet\Policies\Microsoft\FeatureManagement\Overrides\140377743`
+2. If the key value is 1, set it to 0 and restart the node (this disables PR 13179278 which is causing delete LB RPC calls from KubeProxy to fail with Invalid IP Error - ICM: 719903780)
+3. Start a continuous monitoring loop with 10-second intervals
+4. Clean up any stale LB DSR rules found
+
+**Note:** This approach fixes issues on a single node. If the issue is widespread across the cluster, deploy the solution using a DaemonSet:
+
+```powershell
+kubectl create -f cleanup-stale-lb-rules.yaml
+```
+
+This will run the mitigation script as HPC pods on all affected nodes.
+
+### Configuration
+
+You can modify these parameters at the top of the script:
+
+- **`$groups`**: VFP groups to monitor (default: `LB_DSR_IPv4_OUT`, `LB_DSR_IPv6_OUT`)
+- **`$refreshIntervalSeconds`**: Time between cleanup iterations (default: 10 seconds)
+
+## How It Works
+
+### 1. Registry Check
+The script first ensures the feature flag registry key (140377743) is set to 0. If not, it sets the value and restarts the node.
+
+### 2. Endpoint Collection
+- Retrieves all HNS policies
+- Extracts endpoint references
+- Builds a dictionary of valid endpoint IP addresses
+
+### 3. Rule Validation
+For each VFP port and LB DSR group:
+- Lists all rules in the `LB_DSR` layer
+- Extracts DIP (Destination IP) ranges from each rule
+- Compares DIPs against the valid endpoint dictionary
+
+### 4. Cleanup
+- Rules with DIPs not found in active endpoints are flagged as stale
+- Stale rules are automatically deleted using `vfpctrl /remove-rule`
+
+## Output Examples
+
+### Healthy State
+```
+All DIP ranges are present in the dictionary.
+```
+
+### Stale Rules Detected
+```
+Missing DIP ranges:
+ - 10.244.0.25
+ - fdf5:5d67:b9ce:b28f::13f
+Deleting rule : ruleId: ABC123, port: Port1, group: LB_DSR_IPv4_OUT
+```
+
+## Monitoring
+
+The script provides color-coded output:
+- **Green**: Healthy state, all rules valid
+- **Yellow**: Configuration changes or rule deletion in progress
+- **Red**: Stale rules detected
+- **Cyan**: Status updates and iteration markers
+
+## Important Notes
+
+- The script runs indefinitely until manually stopped (Ctrl+C)
+- Node restart may occur on first run if registry configuration is incorrect
+- Ensure no legitimate endpoint updates are in progress during cleanup to avoid false positives
+- The script requires elevated privileges to modify VFP rules and registry settings
+
+## Troubleshooting
+
+### Script doesn't detect stale rules
+- Verify VFP and HNS are functioning correctly
+- Check that `vfpctrl.exe` is accessible in the system PATH
+- Ensure HNS endpoints are properly registered
+
+### Node restarts unexpectedly
+- This is expected behavior if the registry key is not set to 0
+- After restart, the script will continue normal operation
+
+### Permission errors
+- Run PowerShell as Administrator
+- Verify account has rights to modify VFP rules and registry
+
+## Related Documentation
+
+- [VFP Documentation](../../helper/VFP.psm1)
+- [HNS Module](../HNS/)
+- [Network Health Monitoring](../../networkhealth/)
+
+## Support
+
+For issues or questions, please refer to the main repository documentation or open an issue.

scripts/mitigation-scripts/stale-lb-dsr-rules/cleanup-stale-lb-rules.ps1

@@ -0,0 +1,102 @@
+$groups = @("LB_DSR_IPv4_OUT", "LB_DSR_IPv6_OUT")
+$refreshIntervalSeconds = 10
+
+function Get-DipRangesFromRuleText {
+    param([string[]]$RuleText)
+
+    $collect = $false
+    $dips = @()
+
+    foreach ($line in $RuleText) {
+
+        # Detect beginning of DIP Range block
+        if ($line -match "DIP Range") {
+            $collect = $true
+            continue
+        }
+
+        # Stop when FlagsEx or another header appears
+        if ($collect -and $line -match "FlagsEx") {
+            break
+        }
+
+        # Process lines like:
+        # { 10.244.0.25 : 53 }
+        # { fdf5:5d67:b9ce:b28f::13f : 4445 }
+        if ($collect -and $line.Trim().StartsWith("{")) {
+
+            # Remove surrounding { } then trim
+            $clean = $line.Trim().Trim('{','}').Trim()
+            # Use regex to extract IP before last " : "
+            if ($clean -match '(.+)\s*:\s*\d+$') {
+                $ip = $matches[1].Trim()
+                $dips += $ip
+            }
+        }
+    }
+
+    return $dips
+}
+
+$regKeyVal = (Get-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Policies\Microsoft\FeatureManagement\Overrides" -Name 140377743).140377743
+if ($regKeyVal -eq 1) {
+    Write-Host "Registry keys are not zero. Setting reg key to 0 and restarting the node." -ForegroundColor Yellow
+    Set-ItemProperty -Path "HKLM:\SYSTEM\CurrentControlSet\Policies\Microsoft\FeatureManagement\Overrides" -Name 140377743 -Value 0 -Type DWORD
+    Restart-Computer -Force
+    Start-Sleep -Seconds 30
+} else {
+    Write-Host "Registry keys are zero. Continuing the script." -ForegroundColor Green
+}
+
+While($true) {
+    Write-Host "Waiting for $refreshIntervalSeconds seconds for the next iteration..." -ForegroundColor Cyan
+    Start-Sleep -Seconds $refreshIntervalSeconds
+    Write-Host "Starting new iteration to check for stale LB DSR rules..." -ForegroundColor Cyan
+    $dictDstIPs = @{}
+
+    $policies = Get-HnsPolicyList
+
+    $endpointIds = $policies.References |
+        Where-Object { $_ -like "/endpoints/*" } |
+        ForEach-Object { ($_ -split "/")[-1] } |
+        Sort-Object -Unique
+
+    $endpointIds | ForEach-Object {
+        $ipAddress = (Get-HnsEndpoint -Id $_).IPAddress
+        if ($ipAddress -ne $null) {
+            $dictDstIPs[$ipAddress] = $true
+        }
+        $ipv6Address = (Get-HnsEndpoint -Id $_).IPv6Address
+        if ($ipv6Address -ne $null) {
+            $dictDstIPs[$ipv6Address] = $true
+        }
+    }
+
+    $ports = (vfpctrl.exe /list-vmswitch-port /format 1 | ConvertFrom-Json).Ports.Name
+    foreach ($port in $ports) {
+        foreach ($group in $groups) {
+            $rules = (vfpctrl /port $port /layer LB_DSR /group $group /list-rule /format 1 | ConvertFrom-Json).Rules
+            foreach ($rule in $rules) {
+                $ruleId = $rule.Id
+                $ruleText = vfpctrl /get-rule-info /port $port /layer LB_DSR /group $group /rule $ruleId 2>&1
+                if (-not $ruleText) {
+                    Write-Host "No output from vfpctrl"
+                    continue
+                }
+
+                $dips = Get-DipRangesFromRuleText -RuleText $ruleText
+                # Check which DIPs are missing in the dictionary
+                $missingDIPs = $dips | Where-Object { -not $dictDstIPs.ContainsKey($_) }
+
+                if ($missingDIPs.Count -eq 0) {
+                    Write-Host "All DIP ranges are present in the dictionary." -ForegroundColor Green
+                } else {
+                    Write-Host "Missing DIP ranges:" -ForegroundColor Red
+                    $missingDIPs | ForEach-Object { Write-Host " - $_" }
+                    Write-Host "Deleting rule : ruleId: $ruleId, port: $port, group: $group" -ForegroundColor Yellow
+                    vfpctrl /remove-rule /port $port /layer LB_DSR /group $group /rule $ruleId
+                }
+            }
+        }
+    }
+}