I installed Mesos on Fedora 25 using sudo yum install mesos and it worked fine.
Starting Mesos master also worked fine, as per instructions here: https://fedoraproject.org/wiki/Changes/ApacheMesos
I am able to see the mesos-master process running, with ps -e | grep mesos yields nothing.
However, if I try to access localhost:5050 then mesos-master dies silently and further ps -e | grep mesos.
Even if I start a local Zookeeper instance and then use the --zk parameter with sudo -u mesos /usr/sbin/mesos-master ..., I get proper registration of the master into Zookeeper, but accessing the master via port 5050 still makes it die.
A run with INFO-level logging shows that mesos-master is dying with 'Segmentation fault'.
I1222 09:10:01.897171 3537 replica.cpp:658] Replica received learned notice for position 32
I1222 09:10:01.897917 3537 leveldb.cpp:343] Persisting action (18 bytes) to leveldb took 691597ns
I1222 09:10:01.898047 3537 leveldb.cpp:401] Deleting ~2 keys from leveldb took 24314ns
I1222 09:10:01.898114 3537 replica.cpp:679] Persisted action at 32
I1222 09:10:01.898135 3537 replica.cpp:664] Replica learned TRUNCATE action at position 32
*** Aborted at 1482397807 (unix time) try "date -d @1482397807" if you are using GNU date ***
PC: @ 0x1 (unknown)
Segmentation fault
The crash takes place even if I try to connect a mesos-slave to the master via zookeeper - which tells that the problem is not with the zk/master/slave setup but with the fact that the master responding to port 5050 makes it crash.
fedoraproject.org states that the Mesos RPM was created for Fedora 21 and I am running on Fedora 25.
Wasn't the RPM supposed to be backward compatible, considering that it is present in Fedora 25 repositories?