Ok, after some back and forth i managed to establish that the bug is in
file

/opt/omd/versions/default/skel/etc/shinken/shinken-specific.d/module_livestatus.cfg

the mongologs module has hardcoded port of 27017, and on every next site
created omd setup utility would shift the mongodb port by 1 with each new
site, in this case to 27018. And the config file would not be updated with
the correct port.

This caused lockups of the broker service and prevented creation of
aforementioned socket, if i only started one of the sites, but not the
first one.

This is on centos 6.4, 64bit, with rpm from omd repository.

I also had to muck a bit with the check_mk config file
(check_submission='pipe') and symlink tmp/shinken/shinken-apache.cfg to
tmp/nagios/nagios.cfg after starting the site to keep the check_mk tool
working correctly.

I still have some issues with check_mk multisite web interface :

Livestatus problem: Unhandled exception: 452: Completely invalid GET
request 'GET log Columns: log_state_type log_plugin_output log_state
log_lineno host_name service_description log_time log_type Filter: log_time
>= 1364892505 Filter: class = 1 Filter: class = 3 Or: 2 Limit: 1001
Localtime: 1364906905 OutputFormat: python KeepAlive: on ResponseHeader:
fixed16'

but list of services works correctly, and all interfaces are now receiving
livestatus data so at least there's that.


2013/3/29 marcin kowalski <[email protected]>

> I encountered an odd bug in OMD 0.56 setup, not sure if this is known or a
> normal way it operates.
>
> Steps to reproduce.
>
> - Setup a testsite with nagios as core.
> - Enable livestatus TCP socket option.
> - setup the host system to run mk agent
> - Start the site ;
> - use check_mk to add and register localhost to nagios (check_mk -I
> localhost ; check_mk -O localhost)
> - Verify correct operation through thruk/check_mk_multisite web interface.
>
> - Stop the site.
>
> - Switch to shinken as core.
>
> - Start the site.
>
> Socket tmp/run/live exists only when site is stopped, regardless of tmpfs
> settings or whether you erase the tmp/ completely before starting the site.
> This means that check_mk_multisite and thruk are completely useless and
> report problems about lack of socket or no available backends.
>
> After issuing omd stop, the socket reappears. I had no luck tracing what
> restores it, i only know that omd recreates /tmp contents on startup and
> removes stale sockets.
>
> Disabling the TCP socket option for livestatus makes it appear correctly
> after omd start with shinken as core. However passive check_mk results no
> longer update in any of the interfaces (except for the one service with
> active check). I can see all the services in shinken or other interfaces,
> but only the one with active checks updates - the rest slowly becomes
> outdated.
>
> Switching back to nagios, regardless of TCP socket settings restores
> proper operation.
>
> Is there any way to make those two work together?
>
_______________________________________________
omd-users mailing list
[email protected]
http://lists.mathias-kettner.de/mailman/listinfo/omd-users

Reply via email to