Calling set_keepalive on the cluster object breaks things

[pjsg]

I suspect that the set_keepalive method ought to be stubbed to do nothing -- and it certainly shouldn't get forwarded to the actual redis connection.

[skrussell]

Seems like we might be hitting the same issue: 2021/07/07 16:36:19 [error] 1759615#1759615: *1 [lua] rediscluster.lua:88: release_connection(): set keepalive failed:closed