Ah! The problem was I was not submitting to the leading master. When I changed the --master IP to the leading master the task got submitted. I changed the command from sleep to python.
root@cnlvr01r08s1:~# GLOG_v=1 LIBPROCESS_IP=10.2.0.219 mesos-execute --master=10.2.1.118:5050 --name="cluster-test" --command="python -m SimpleHTTPServer" WARNING: Logging before InitGoogleLogging() is written to STDERR I1228 08:17:38.477524 42862 process.cpp:965] libprocess is initialized on 10.2.0.219:46533 for 48 cpus I1228 08:17:38.477645 42862 logging.cpp:198] Logging to STDERR I1228 08:17:38.478806 42862 sched.cpp:166] Version: 0.26.2 I1228 08:17:38.482700 42863 sched.cpp:264] New master detected at master@10.2.1.118:5050 I1228 08:17:38.482929 42863 sched.cpp:274] No credentials provided. Attempting to register without authentication I1228 08:17:38.482952 42863 sched.cpp:716] Sending SUBSCRIBE call to master@10.2.1.118:5050 I1228 08:17:38.483043 42863 sched.cpp:749] Will retry registration in 1.442011578secs if necessary I1228 08:17:38.483903 42894 sched.cpp:643] Framework registered with e4b204c0-4d3a-44c9-880c-03bd0063fe26-0013 Framework registered with e4b204c0-4d3a-44c9-880c-03bd0063fe26-0013 I1228 08:17:38.483963 42894 sched.cpp:657] Scheduler::registered took 29377ns I1228 08:17:38.487526 42882 resources.cpp:472] Parsing resources as JSON failed: cpus:1;mem:128 Trying semicolon-delimited string format instead task cluster-test submitted to slave e4b204c0-4d3a-44c9-880c-03bd0063fe26-S3012 I1228 08:17:38.487763 42882 sched.cpp:813] Scheduler::resourceOffers took 266521ns Received status update TASK_RUNNING for task cluster-test I1228 08:17:38.595126 42895 sched.cpp:921] Scheduler::statusUpdate took 37167ns I1228 08:17:43.826679 42871 sched.cpp:813] Scheduler::resourceOffers took 41784ns Before I got the task scheduled I the following in the log of the master that was not leading: I1228 08:32:10.493804 36745 master.cpp:1193] Dropping 'mesos.scheduler.Call' message since not elected yet I1228 08:32:10.724387 36745 master.cpp:1193] Dropping 'mesos.scheduler.Call' message since not elected yet Perhaps mesos-execute should check if the master is leading first? Or is this already fixed in versions after 0.26.0?
