Details
-
Improvement
-
Resolution: Unresolved
-
Major
-
None
-
None
-
None
-
2
Description
So that we can properly support HTTP fallback we should do the initial bootstrap slightly differently to how we bootstrap steady state connections. We should try each node in the bootstrap list in serial and a get cluster config request failing should be considered fatal to the attempt. If bootstrap against all the nodes fails then we should attempt to fetch a config over http. If that fails then back to attempting memcached on the connections one by one.