-
Type: Task
-
Resolution: Done
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
Labels:None
When a sharded cluster is using three different config servers, the only way that the cluster knows that one of them is down is through the --configdb string. If someone changes the --configdb string used by the 'mongos' processes, then the cluster can go into read/write mode and potentially lose all of the data inserted.
Docs should be explicit about not changing the --configdb string in case of config server downtime.