As discussed previously on ceph-devel, cherrypy on centos is a relic, and was monkey patched at least once, for ssl issues [0]. There's another issue that has popped up with cherrypy, this time with ipv6 in rook. I think it is probably better to solve this on the rook side, but wanted to put it up for discussion: By default the dashboard binds cherrypy server to any interface (i.e. "::"). The version of cherrypy on centos throws an exception [1] when the host doesn't have ipv6 loopback support configured (which seems to be the common case in the containers running on kubernetes). Newer versions of cherrypy just log a warning. For various reasons Rook would benefit from binding to "::", but cannot because of this bug. Since Kubernetes doesn't yet support ipv6 [2] this is a big deal--we can bind to 0.0.0.0, but it does mean that Rook has to later detect ipv6 support and reconfigure itself to use "::". The big question for ceph-devel is... is ipv6 so widely supported now by default that monkey patching cherrypy to remove this check/bug would _only ever_ benefit Rook? And if so, perhaps we should just handle the fix there. [0]: https://github.com/ceph/ceph/commit/ce92ddf4 [1]: http://tracker.ceph.com/issues/37651 [2]: https://github.com/kubernetes/enhancements/issues/508
