This bug is reported from user list, the document should be updated by removing multiple masters when slave start up.
/usr/sbin/mesos-slave --master=IP1:5050,IP2:5050,IP3:5050 …. --credential …
Only if IP1 is the leader, the slave can register to master successfully, Or it will register fail.
Slave log like this:
Creating new client SASL connection
Authentication timed out
Failed to authenticate with master firstname.lastname@example.org:5050: Authentication discarded
Authenticating with master email@example.com:5050
Using default CRAM-MD5 authenticatee
Is this a bug?Or it is designed like this.
BTW: --master:zk://xxxxxxx work well.