stop and start hdfs-mesos and multiple hdfs-mesos frameworks

[sonoroot]

Hello, I'd like to have some suggestions on how to achieve or getting close to the following: 1 - How do you completely stop and start the hdfs-mesos framework, maintaining a consistent HDFS cluster? (for example, using a scheduler that pins the namenodes and journal nodes to specific slaves ?) I'm wondering what happens if the entire mesos cluster goes down (i.e. datacenter power failure) will I be able to recover the HDFS cluster?

2 - Is it possible to run multiple hdfs-mesos frameworks? (for example customizing data dirs locations, framework names, TCP ports of journalnodes, namenodes and datanodes etc. for each hdfs cluster)

Thank you!

[elingg]

1. Regarding backup and recovery, we do have this work on our roadmap. In addition, there is a distcp function that copies from an active HDFS cluster to another one.

2. Multitenancy is not yet possible, but it's something farther out on the roadmap. PR's are welcome.