A Quick Thought About Mesos-DNS
A colleague recently pointed me to the recent Mesosphere announcement of Mesos-DNS, a DNS-based service discovery mechanism for Apache Mesos clusters. A comment made in the announcement got me thinking, and I wanted to briefly share my thoughts.
The comment that got me thinking was this:
Mesos-DNS is simple and stateless. Unlike Consul and SkyDNS, it does not require consensus mechanisms, persistent storage, or a replicated log.
If you’ve been following along here on my site, you know that I’ve written about Consul before (see here), and I have more Consul content planned. I’m reasonably familiar with Consul’s architecture and requirements. Likewise, although I haven’t specifically written about SkyDNS, it’s based on etcd, which I have talked about (see here). The Mesos-DNS article seems to imply that Mesos-DNS is somehow less complex than either of these two solutions because it doesn’t require consensus mechanisms, persistent storage, or a replicated log.
However, in my mind that’s a misleading statement. Yes, Consul does require a consensus mechanism (it uses Raft, like etcd). SkyDNS (as I understand it, at least) simply leverages etcd, so technically SkyDNS itself doesn’t require a consensus mechanism. And the assertion that SkyDNS itself Continue reading