Investigate whether we can use "reconcileTasks" method rather than custom healthcheck message
The healthcheck sends a custom framework message to request a status update. We might be able to use the mesos method reconcileTasks().
wait for refactoring #327
FWIW, We experienced an issue where an executor task went lost while the scheduler was also down, and we were unable to recover gracefully from this. I assume this is due to the executor healthcheck being done by the scheduler directly and not by Mesos.
@justinclayton Thanks Justin. We've just had a similar issue reported with #550. We can't decide whether this is our responsibility. I.e. whether we should add more code to work around the fact that Mesos can't handle this failure.