Backend Guy
"Sup."
https://twitter.com/francesc/status/487412202932936704
representing 1,631,392 lines of code
maintained by a very large development team
with increasing Y-O-Y commits
starting with its first commit in June, 2014
ending with its most recent commit about 1 month ago
https://www.openhub.net/p/kubernetes
http://kubernetes.io/docs/api-reference/v1/operations/
https://github.com/kubernetes/kubernetes/issues/15128
https://github.com/kubernetes/kubernetes/issues/1362
Persistence clusters really don't like nodes disappearing/reappearing.
Or.. at least redis sentinels don't
Neither do memcached farms
Stable DNS identity only solves part of the problem, but not when dealing directly with IPs
$ kubectl get pods --all-namespaces \ -o yaml | grep 10.244.18.6 $
Hunting for an IP across all pods, across all namespaces
/data # redis-cli -h 10.244.18.6 \
-p 26379 info | \
grep "redis_version"
redis_version:3.2.1
Surprise!
/data # redis-cli -h 10.244.18.6 -p 26379 10.244.18.6:26379> shutdown not connected>
Kill it with Fire
Well, just with kube-up.sh
// Jumpbox
$ sudo find / -name 'ca.key'
$ <nothing>