Elizabeth Lingg

Sounds awesome @knuckolls. There should be a resource check when launching them to make sure that 2/3 journal nodes have space for the NN's to colocate on them.

cc @adam-mesos

thanks @adam-mesos, I think upgrading the version of Hadoop would be a specific scenario in the upgrade guide. There's upgrading the framework version and then there's upgrading the version of...

correct @tangzhankun, the answer is no currently. Thanks for bringing up this issue @teamsoo! Constraints in HDFS-Mesos may be quite useful

Zookeeper storage with an in memory cache may be a good solution for this. In fact, refactoring of persistent state to use an in memory cache would be ideal.

hi @tangzhankun, With the current implementation, if a node dies it has 1.5 minutes to recover (this time is configurable). If the scheduler fails over and a node has died...

@abhay-agarwal , yes this will be fixed when there is a configurable number a DN's. A workaround is that when the slave dies, a new slave gets spin up and...

Corrected Steps: 1. the scheduler's run() will execute and then callback registered() or reregistered() will be called. 2. driver.reconcileTasks() will be called which will send status updates for all running...

hi @tangzhankun, yes your description of the issue is correct.

Correct, not serious, but we still want to get rid of these bugs as well!