DC/OS and k8s have some overlaps, but I wouldn't characterize the two as direct competitors.
Majority of DC/OS users are running distributed data services like Spark, Kafka, Elastic, Cassandra, etc. And they're doing this for at least three reasons 1) Pooling these distributed systems instead of running them in separate clusters, 2) The automation for scaling and high availability that Mesos frameworks enable, and 3) Being able to run these data services in any cloud they choose - Google, Azure, AWS, or their own. The 2016 Mesos Survey (n=500) has more details.
If you are squarely focused on container orchestration, 62% of DC/OS and Mesos community runs containers in production. Average size of their environments is 200+ nodes. A similar survey of the K8s community showed rough average K8s environment is around 20 nodes.
