dependency on cassandra
joeblew99 opened this issue · 3 comments
Could it makes sense to use etc or consul for this ?
It makes deployment etc easier with a 100% golang stack.
Thanks for the suggestion. Right now running metadata on etcd/Consul/zookeeper isn't on our priority list. That said, we have designed metadata access layer behind a Thrift interface, so that it is not cumbersome to support different distributed database. We welcome community contribution to make this happen!
One of the main reasons for using Cassandra was to stay available during network partitions. Any metadata solution that requires quorum to operate is going to have lower availability. I still think it would be useful to support alternative metadata implementations, but I'm not sure how many users would prefer easier deployment over higher availability.
I think it's better to remove external dependence, easier deployment is welcomed.