High Availability Guidance when Setting Up a Local Edition Cluster on Swarm Deployments
Health Check Issues
Continuous health check is turned ON for Swarm deployments in API Management - Local Edition 5.4. So, all instances will attempt to ping their binding IPs and if the ping fails, will attempt to re-establish their bindings automatically. No manual rebinding is necessary for Local Edition 5.4. Swarm deployments.
Binding Issues
This enhancement ensures that services and their downstream dependencies (for example, Log) are automatically rebalanced to ensure load distribution. Any log service failure causes a reshuffling of bindings in all upstream instances. With Local Edition 5.4 Swarm deployments, a respinning of log service results in all upstream components (such as TM, Cache, SQL) automatically detecting the failure via health-check and rebind happens once a new log service is available.
During this interval, the TM continues to serve traffic and log events are collected locally on the TM. They will be forwarded to the log service, once the log service comes up and the TM (Fluent-bit process) binds to it automatically.
No manual intervention or reconfiguration is necessary in Local Edition 5.4 Swarm deployments for the binding issues.