HA Architecture for LAMs

The Moogsoft AIOps HA architecture provides increased resilience against LAM and server restarts by caching ingested data to the disk. It requires installing a local RabbitMQ cluster which is used by LAMs for publishing.

A remote caching LAM, located next to the Core role, connects to the local RabbitMQ cluster, picks the events from the queue and publishes them to the central RabbitMQ cluster for Moogfarmd to process.

If no caching LAM is available to consume the events from the local RabbitMQ cluster, the data is cached to disk until the server runs out of memory.

HA architecture

This architecture is recommended for hybrid installations, where the core processing is located in the cloud and LAMs are on-premise, or for a full on-premise configuration where LAMs are housed remotely to the core components.

LA_LAMwithcache.png

Polling LAMs run in an active / passive mode and must connect to a local database in order to negotiate their state. This requires a local MySQL instance that runs with master / master replication.

Polling_LAM_HA_diagram.png

Installation steps

If you are setting up the Store and Forward architecture, perform the following steps:

Otherwise, perform the standard installation steps: