Re: URL content checking [Vote]

[Michael Shigorin]

PreScriptum: external (runnable) tests should do?

On Thu, Dec 02, 2004 at 06:20:12PM +0100, Jan-Henrik Haukeland wrote:
> Nothing in the proposed URL function contradict this
> philosophy. No plugin nor external libs are required to realize
> this functionality. 

That's the entire point of my post: eventually such a creeping
expansion tends to result in some bloated mess -- that's happened
all over...

> The point is that monit already have support for url-testing.
> So what we actually are doing is to lift up existing
> functionality into the control file and wrap it in a new
> setting.  A bit of "glue" coding is necessary but nothing
> fancy.

These add up.  You could also do RPM base lookups to determine
which packages (should) contain the files with broken checksums,
and for at least part of monit users it would be beneficial.
At the same time, having librpm linked in doesn't help with being
bulletproof as ld will blow up with e.g. libdb4.so being broken.

> >I may be wrong but a quick perl, ruby, python, heck -- shell
> >script to GET or lynx -dump the page(s) and some grepper would
> >do this particular trick.
> You are not wrong, but does that imply that it is wrong to
> implement it in monit? I think not, actually monit is great for
> stuff like this.

Actually it's not a platform.  If it was allowing test code to be
external (and being able to act differently on varying error
codes and inability to run the test) -- it would be.

That's what I've been talking about later.

> >So all in all, is monit the targeted system-level bulletproof
> >tool focusing on local processes or is it aiming to become a
> >distributed infrastructure wielding dozens of plugins?
> ``Every program attempts to expand until it can read mail.
> Those programs which cannot so expand are replaced by ones
> which can.''

Sure! :-)

> But seriously, monit already does "bulletproof system
> monitoring" and  if you  only need that, stay with monit
> version 4.4. Monit need to evolve if it should continue to be
> alive and an active open-source project. If there is nothing
> more to do, the project will die.

Oh no, that's a hype! :-)  Breaking existing functionality,
"bulletproofness" being one of the critical features, is a major
bug for such a software.

You know, I've tried to vote monit into being more fundamental
feature of ALT Linux with thorough support by server packages
(monitrc.d/ files) -- but the final word was that "it's too
complex to be core component".

As with m/monit, sometimes it's just good to split things.

> So to stay alive we need to add stuff to monit and continuously
> refactor and refine it.

That's the question: features and security/stability always fight
for developers' time and users' community.

> It's a question about choosing the right direction for the
> project.

Exactly.  As I'm no C developer (short of number-crunching parts
of a B.Sc. project somewhere in the past) and hardly a proper
shell/Ruby developer, it's hard for me to contribute with code.

But with some analysis over projects I've helped architecturing
(mobile content storage/delivery system -- maybe I've mentioned
it already since we've made it quite modular and it really helped
when things started to change) -- or just spectated -- it seems
that putting flies together with sausages doesn't really work.

> Right now it seems that monit is evolving into becoming a
> monitoring swish-army knife which doesn't sound to bad in my
> book.

You know, they're splitting one such knife (Mozilla) into
xulrunner (a platform) and a bunch of applications over it. :)

I believe that the simplest approach to this would be adding
support for external test programs -- just like with mrtg, you
can then write (and contribute) any script or compiled code.

> I know it comes from a genuine interest in monit so I do very
> much appreciate your points of view and comments. But don't
> expect us to follow them all the time :)

I don't expect it at all, I have no vote without SLOC.  Really.

> Nice, BTW I hope everything is okay and that things will settle
> in your country.

Thanks, Jan.  It's a couple of kilometers away, people under the
rain...

-- 
 ---- WBR, Michael Shigorin <address@hidden>
  ------ Linux.Kiev http://www.linux.kiev.ua/