Re: Host connection test and alert/exec

[Jan-Henrik Haukeland]

Andreas Rust <address@hidden> writes:

> the idea of resolving this is too good to be true ofcourse. :)

Of course :) it's not a bad idea though.

> It appears that the term "within X cycles" is only available on
> resource testing and service timeout. the [timeout] option on the
> connection testing is for the time monit should keep trying to connect.
> A very nice option aswell, I have to say :)
>
> Unfortunately it doesn't work this way.
> Maybe an idea to add this option to the parser here ?

Maybe, but from your original mail:

> Now, since remote hosts may disappear due to the line being cut or
> something alike, I'd like to send my alert on a timeout (hence, on
> several failed tests), similar to normal process checking.  Means,
> that some X connection failures are ok, alert me on X+1.

I don't understand why it is not interesting to know if the line was
cut right away and get and alert? If you are thinking about that you
will get an alert every time monit checks the remote host and it's
still down, then this is fixed in the upcomming 4.1 release: If the
host is down monit will now only send you one alert and as long as the
host is down you will not get anymore alerts. But if the host got up
and then down again later you will get a new alert (and so on). (Check
out 4.1 beta release).

> Since HOST checking only allows alert, restart, stop, exec and
> unmonitor, I figured I'd have to use an exec script to simulate this
> functionality.

Unfortunately yes, there is no functionality in monit now to do what
you want. But using an exec -> script for this should be fairly easy.

-- 
Jan-Henrik Haukeland