This is an umbrella issue for all the KV store related failures to help addressing those:

• persistent failure of atomic kv operations until docker is restarted #576 Persistent Failure of Atomic Operation until docker is restarted
• Removing containers doesn't remove entries in overlay network. #573 Removing containers does not remove entries in overlay network
• bit sequence handler does not initialize with zookeeper backend. #523 Bit sequence handler does not initialize with zookeeper backend
• --kv-store parameter does not use URI format #438 KV store parameter does not use URI format
• Restarting the KV store causes Docker Daemons to peg the CPU to 100% #408 Restarting the KV store causes the CPU to peg at 100%
• Etcd can't be used a datastore on multihost #392 Etcd can't be used as a datastore on multi-host
• Zookeeper, etcd support multiple IPs but datastore.newClient only takes one. #507 Zookeeper, etcd support multiple IPs but datastore.NewClient only takes one
• docker network create fails with "failed to allocate vxlan id: no bit available" #559 docker network create fails with "failed to allocate vxlan id: no bit available"