A bunch of usage issues with cctl
roopakparikh opened this issue · comments
a) Calling ./cctl status, just returns the following:
./cctl status Status called
If unimplemented we should state so.
b) Calling delete machine with no arguments has a very weird message
./cctl delete machine 2018/09/18 04:29:46 Unable to get machine "": more than one object matched gvr {cluster.k8s.io v1alpha1 machines}, ns: "default" name: ""
c) Invalid value of routerID causes havoc when during machine creation, I created a cluster with routerID '501' which went through but caused the machine create to hang forever.
d) There is no way to create a cluster without VRRP (not even a single node cluster).
e) etcd snapshot is a different operation than backup (and similarly restore portion).
Thanks for filing these @roopakparikh
(a) We should probably remove all of the commands that have not been implemented. Creating a separate issue for that (issue #119 )
(b) That doesn't look right. Will look into this
(c) Dup-issue. Issue #53
(d) This is by design. We want to be able to scale up from a single master cluster to a multi-master cluster. Starting with a non-VIP based cluster prevents this operation. Although we should be able to use a VIP but not provision keepalived. This is a feature add. Will create a separate issue for that.
(e) Yes, by design. We have two pairs here - backup and restore which creates and extracts a backup tar archive from an etcd snapshot and the cluster (cctl) state file. The second one being snapshot etcd and recover etcd which are used for etcd backup/recovery operations.