Merge pull request #104386 from peterpogorski/seed-node-deletion-crp-sfrp

Seed node deletion crp sfrp

Author: v-albemi

articles/service-fabric/service-fabric-azure-resource-manager-guardrails.md

@@ -6,7 +6,7 @@ documentationcenter: .net
 author: peterpogorski
 
 ms.topic: conceptual
-ms.date: 10/30/2019
+ms.date: 02/13/2020
 ms.author: pepogors
 ---
 # Service Fabric guardrails 
@@ -55,12 +55,27 @@ The following section contains an example of a durability mismatch between the V
 * Virtual Machine Scale Set durability does not match the target Service Fabric Node Type durability level
 * Virtual Machine Scale Set durability does match the current Service Fabric durability level or the target Service Fabric Node Type durability level 
 
-
 ### Mitigation
 To fix a durability mismatch, which is indicated by any of the above error messages:
 1. Update the durability level in either the Virtual Machine Scale Set extension or Service Fabric Node Type section of the Azure Resource Manager template to ensure that the values match.
 2. Redeploy the Azure Resource Manager template with the updated values.
 
+
+## Seed node deletion 
+### Overview
+A Service Fabric cluster has a [reliability tier](https://docs.microsoft.com/azure/service-fabric/service-fabric-cluster-capacity#the-reliability-characteristics-of-the-cluster) property which is used to determine the number of replicas of system services that run on the primary node type of the cluster. The number of required replicas will determine the minimum number of nodes that must be maintained in the primary node type of the cluster. If the number of nodes in the primary node type goes below the required minimum for the reliability tier, the cluster will become unstable.  
+
+### Error messages 
+Seed node removal operation has been detected, and will be rejected. 
+* This operation would result in only {0} potential seed nodes to remain in the cluster, while {1} are needed as a minimum.
+* Removing {0} seed nodes out of {1} would result in the cluster going down due to loss of seed node quorum. Maximum number of seed nodes that can be removed at a time is {2}.
+ 
+### Mitigation 
+Ensure that your primary node type has enough Virtual Machines for the reliability specified on your cluster. You will not be able to remove a Virtual Machine if it will bring the Virtual Machine Scale Set below the minimum number of nodes for the given reliability tier.
+* If the reliability tier is correctly specified, make sure that you have enough nodes in the primary node type as needed for the reliability tier. 
+* If the reliability tier is incorrect, initiate a change on the Service Fabric resource to lower the reliability level first before initiating any Virtual Machine Scale Set operations, and wait for it to complete.
+* If the reliability tier is Bronze, please follow these [steps](https://docs.microsoft.com/azure/service-fabric/service-fabric-cluster-scale-up-down#manually-remove-vms-from-a-node-typevirtual-machine-scale-set) to scale down your cluster gracefully.
+
 ## Next steps
 * Create a cluster on VMs or computers running Windows Server: [Service Fabric cluster creation for Windows Server](service-fabric-cluster-creation-for-windows-server.md)
 * Create a cluster on VMs or computers running Linux: [Create a Linux cluster](service-fabric-cluster-creation-via-portal.md)