operational update

Last night, we made a tweak to our implementation last night that broke the c:resolve and yaml renderer.  We reverted the change to get this back to operational this morning.

We've resolved these problems and will be redeploying the change soon (after a tad more testing).  We are improving our database connection manager.

1.  Add the ability to read from all database hosts.  Currently, we only read from the primary database server and the other servers (there for redundancy) are not used at all.  With this change, these other hosts will be used as READ-ONLY hosts (as long as they are current with their data) so that incoming requests can be spread amongst more servers.  This should give us some more performance.

2.  Add failure recovery when the primary database host changes.  Currently, connections established will simply fail until the redundancy kicks into gear and a new PRIMARY host is available.  With the change, our code will detect this and simply retry for a while until it can make the connection and complete the request.

This change affects both requests coming in through the API as well as updates coming in from the game servers.

-dan

 

These changes have been deployed. 

-dan

 

Thanks again for all the hard work.  The recent changes have been massive and smooth.