Reference for ROMA
ROMA provide acluster replication function which is Hot Standby logic.
This new function copy the data from main cluster to replica cluster as also existing data.
Also ROMA has auto recovering sisytem and other counter measures for cluster error,
but if you use this func, you can revert the cluster by just switching access host of API.
This function automatically keeps redundancy if some node down.
ROMA support redundancy and dispersibility.
And basically redundancy constructed with some servers(not only 1 sever.)
When some nodes down and break this redundancy, surviving instance copy data to another surviving instance by automatically.
So if trouble happened, Redundancy continue to keep without human operation!!
Snapshot has 4 status when ROMA copy a data as backup file. Their status are ":normal", ":flushing", ":flushed" and ":cache cleaning".
This function works as backup function if cluster or server have a problem.
After ROMA copy all data which memory has to tc file, it is copied as backup file. And ROMA guarantee the latest data of tc file
by using the caching file. Also, ROMA guarantee the latest data which ROMA return to client by comparing the clock of "cache file"
and "tc file".
In case of DNS server down & ROMA's routing were made by not IP address but short-hostname, ROMA can NOT access another server's instance.
As a counter plan, ROMA keep DNS information for each instance as a cache file and connect by using this.
So even so DNS server will down, ROMA can keep connect to other server's instance.
These cache will update when activated and deactivated (by use changing command), or instance information changed(ex.join, recover).