First thing I see is that you have a quorum of 2, but only 2 masters. For a quorum of 2, you should have 3 masters. Logic: for an odd number of M masters, a quorum/majority of ceiling(M/2) is necessary to win a vote. For quorum Q, you should launch (2*Q)-1 masters. What you are showing is expected behavior for a slave when the masters keep switching. It would be more valuable to see the log from one of the masters to determine why it shuts down.
On Tue, Jan 27, 2015 at 7:48 PM, xiaokun <xiaokun...@gmail.com> wrote: > I followed the instruction in page > http://mesosphere.com/docs/getting-started/datacenter/install/. > Setup two masters and one slave. And quorum value is "2". Configured ip > addresses in hostname files separately. > Here is the log from slave node, > I0127 22:37:26.762953 1966 slave.cpp:627] No credentials provided. > Attempting to register without authentication > I0127 22:37:26.762985 1966 slave.cpp:638] Detecting new master > I0127 22:37:26.763022 1966 status_update_manager.cpp:171] Pausing sending > status updates > I0127 22:38:06.683840 1962 slave.cpp:3321] Current usage 16.98%. Max > allowed age: 5.111732713224155days > I0127 22:38:26.986556 1966 slave.cpp:2623] master@10.27.17.135:5050 > exited > W0127 22:38:26.986675 1966 slave.cpp:2626] Master disconnected! Waiting > for a new master to be elected > I0127 22:38:34.909605 1963 detector.cpp:138] Detected a new leader: > (id='2028') > I0127 22:38:34.909811 1963 group.cpp:659] Trying to get > '/mesos/info_0000002028' in ZooKeeper > I0127 22:38:34.910909 1963 detector.cpp:433] A new leading master (UPID= > master@10.27.16.214:5050) is detected > I0127 22:38:34.910989 1963 slave.cpp:602] New master detected at > master@10.27.16.214:5050 > I0127 22:38:34.911113 1963 slave.cpp:627] No credentials provided. > Attempting to register without authentication > I0127 22:38:34.911144 1963 slave.cpp:638] Detecting new master > I0127 22:38:34.911183 1963 status_update_manager.cpp:171] Pausing sending > status updates > I0127 22:39:06.684526 1964 slave.cpp:3321] Current usage 16.98%. Max > allowed age: 5.111731773610567days > I0127 22:39:35.231653 1963 slave.cpp:2623] master@10.27.16.214:5050 > exited > W0127 22:39:35.231869 1963 slave.cpp:2626] Master disconnected! Waiting > for a new master to be elected > I0127 22:39:42.761540 1964 detector.cpp:138] Detected a new leader: > (id='2029') > I0127 22:39:42.761732 1964 group.cpp:659] Trying to get > '/mesos/info_0000002029' in ZooKeeper > I0127 22:39:42.762914 1964 detector.cpp:433] A new leading master (UPID= > master@10.27.17.135:5050) is detected > I0127 22:39:42.762984 1964 slave.cpp:602] New master detected at > master@10.27.17.135:5050 > I0127 22:39:42.763089 1964 slave.cpp:627] No credentials provided. > Attempting to register without authentication > I0127 22:39:42.763118 1964 slave.cpp:638] Detecting new master > I0127 22:39:42.763155 1964 status_update_manager.cpp:171] Pausing sending > status updates > > Whenever the slave try to connect, the master will existed and try to > elect another one. Is there anything wrong with my configuration? > > Thanks, > Xiaokun >