High Availability

High Availability (HA) Overview for Keeper Connection Manager

Overview

Below is a high-level overview of setting up Keeper Connection Manager (KCM) for High Availability (HA). This is an advanced implementation that requires familiarity with load balancing, databases, security groups, VPCs, etc.

HA Approaches

Active-Active

• Multiple KCM servers handle traffic simultaneously.

• Requires a load balancer with session persistence (sticky sessions) to ensure users stay connected to the same KCM instance.

• Provides higher throughput and automatic failover.

Active-Passive

• One primary KCM server is active, with a standby instance ready to take over.

• Failover occurs only if the active server becomes unavailable.

• Simplifies troubleshooting but does not improve overall capacity.

Key Requirements

• External Database – All KCM instances must share a single database (e.g., MySQL, PostgreSQL) to ensure consistency.

• Load Balancing – Required for both HA models. Use a load balancer with sticky sessions for Active-Active or a failover mechanism for Active-Passive.

• Storage Considerations – Session recordings in KCM are stored on the local disk of the KCM instance. If storage is full, new recordings fail while old ones remain. In a cluster, you may need to consider a shared drive or a cron job to sync the files.

• SSL Termination: By default, the KCM docker compose is configured with NGINX to terminate SSL with either a custom certificate or Let's Encrypt certificate. Based on your load balancer's features, you can decide to terminate the SSL at the load balancer or terminate on the KCM instance.

Setup Steps

1

Deploy Multiple KCM Instances

Ensure that the configuration is identical, using the same docker compose.

2

Deploy Database

• Set up a cloud managed or locally managed database

• Ensure that database backups are enabled

• Migrate KCM data from local DB to shared DB (e.g. using mysqldump)

3

Update KCM Configuration

• Update the KCM instances (e.g. docker-compose.yml) to point to the shared DB

• Configure the listeners on each container based on the load balancer setup

4

Implement HA Model

• Active-Active: Deploy a load balancer with sticky sessions

• Active-Passive: Set up monitoring to detect failures and switch traffic

5

Test & Validate

• Simulate failures, monitor performance, and verify automatic failover