Re: New monit web-interface

[Jan-Henrik Haukeland]

Christian Hopp <address@hidden> writes:

> Sorry for being not very precise.  This display you realized is
> already nice, but I thought of something in a more "per service"
> way.  I like to know how long a service, e.g. apache, is down or up.

Aha, good idea. We'll do, but tomorrow

> The time of next check could be also on a per service base.  If you
> fiddle around with the "every" statement it might be interesting.

Well, maybe. But don't expect all your whishes to come through :)

> I don't know if it might be less annoying for a server if you
> scatter the tests in the cycle (e.g. m tests in n seconds, means n/m
> seconds delay between each service.). 

It's probably better, to avoid the peaks but..

> It's not nice because you have to split up validate().

and you know me, mr. keep it as simple as possible to make it work

> Is the md5sum check done every cycle? 

Yes

> Wouldn't it be enough to check ctime first and if newer then last
> cycle do a md5sum check.

Yes another good idea, but..

> Some server programs might come up to some megs.

The check is pretty fast (for this type of application), aprox 0.07
sec (cpu time) for 2 megs.

> Nice redundant code... at the end of monitor.c is a duplicated...
> | static void start_http() {
> |
> |  monit_http(START_HTTP);
> |
> | }
> ... do they get lost sometimes? (-:


It's a function, modelled after the facade pattern, where we call into
a sub-system. In case we have to do some pre or post fix operations
the entry function provides an abstraction and a unified interface to
the cervlet sub-system. Albeit we don't do any suchs operations here
now (we did before and the need can still arise).

-- 
Jan-Henrik Haukeland