When MMM starts, it checks all agents. If it can not reach any agents, it is supposed to go into passive mode.
Instead, it dies with the following message:
[openquery@mmm2-monitor ~]$ sudo service mysql-mmm-monitor start
Daemon bin: '/usr/sbin/mmmd_mon'
Daemon pid: '/var/run/mmmd_mon.pid'
Starting MMM Monitor daemon: DIED: Network is unreachable at /usr/sbin/mmmd_mon line 60
failed
On top of that, either the message is wrong, or the way this is determined is wrong: when all agents are unreachable, the network is not dead, the agents are merely unreachable :)
There is not really a good way to determine if "the network" is dead, but some options are checking for an ip other then 127.0.0.1 on the machine, checking if any interface is up, etc.
This is mmm 2.0.9
mmmd_mon tries to reach all "ping_ips" defined in the config - if one is reachable, network is considered ok. You should add your switch (or another reliable host) to ping_ips.
Changed the log message for now, I'm not sure why I let mmmd_mon die in this situation, but there must be a reason. ;-)
The description of the "Network check" (section 5.5) is wrong. (unless we change the code).
Someone should try out whether everything works when starting up without network (without the LOGDIE) - maybe we can just remove it.