add multi agent replicas info

Author: guowenatk

platform/configure/config.mdx

@@ -293,7 +293,7 @@ Even if you disable automatic deployment, the platform agent is still required i
 
 ### Install the agent manually {#manual-agent-installation}
 
-To manually install the platform agent:
+Use the following command to manually install the platform agent from Helm releases:
 
 ```bash title="Manual agent installation"
 RELEASE_NAME=vcluster-platform
@@ -308,6 +308,29 @@ helm upgrade $RELEASE_NAME vcluster-platform \
   --set agentOnly=true \
   ${PLATFORM_VERSION:+--version $PLATFORM_VERSION}
 ```
+### Multiple agent replicas deployment
+
+For high availability and scalability, the platform agent can be deployed with multiple replicas. However, certain agent tasks must be executed by only one pod 
+at a time to prevent duplicate actions, race conditions, or data corruption. Other tasks, such as concurrent background processing, can safely run in parallel 
+across replicas.
+
+#### Leader election
+
+Before scaling the platform agent beyond a single replica, you must enable leader election. This ensures that only one pod performs “singleton” tasks while 
+others act as standby or worker agents. Without leader election, multiple pods may perform the same critical operation, leading to inconsistent or 
+conflicting results.
+
+#### Recommended operational practices
+- Enable leader election prior to scaling the agent above one replica.
+- For sensitive operations (e.g., storage migrations or major version upgrades), temporarily scale the agent to a single replica, complete and validate the operation, then scale back up with leader election enabled.
+
+The agent is deployed and managed as a Helm release when the platform connects to a cluster. Configure replicas and leader election settings through Helm values, 
+not manual scaling, to ensure consistent configuration and reliable leader election.
+
+:::warning Avoid manual scaling of agent replicas
+Do not manually scale Deployment replicas outside of Helm. Doing so can start extra pods without the expected flags or config, resulting in no leader 
+being elected, race conditions, or duplicate execution of singleton activities.
+:::
 
 ## Configure the agent {#agent-configuration}