search

inthefabric / Fabric

The collective mind awaits your input.
www.inthefabric.com
Other
5 stars 0 forks source link

Node Monitoring and Retries #12

Open zachkinstner opened 11 years ago

zachkinstner commented 11 years ago

Currently, each API request gets assigned a particular cluster node, and makes its calls to that node. A better approach would be to tell the ApiDataAccess class about all nodes, and allow it to switch between nodes if one becomes unavailable.

This should all be transparent to the users of ApiDataAccess/ApiContext.

Also, the ApiDataAccess class should continually ping each node to keep connections alive and to track node availability.

zachkinstner commented 11 years ago

Seems like something a load balancer should handle. Moving this to "ideas" for now, but it could probably be closed.