Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline
Hi list,
I finish my patch for the host path part. I try to folow the
indentation and the coding style of nagios.
you can find test files with a lot of parent/childs at
http://zegabes.free.fr/nagios/ .
I try to generate a big configuration for service dependencies (4000
dependencies) but it's ok (0.2s). So I don't know if I'vegot to check
it if it's already ok. Someone got problem with this check or it's was
just with host path check?
Ask me if you have a question about my patch, or if it need more work
for include it into a official version of nagios.
Thanks,
Gabes Jean
On Jan 30, 2008 3:21 PM, nap wrote:
>
> On Jan 30, 2008 3:04 PM, Andreas Ericsson wrote:
> >
> > nap wrote:
> > > Hi Ethan,
> > > Hi Everyone,
> > >
> > > I already send this message to the list by another adresse, but it
> > > failed (exchange server...), so I resend it.
> > >
> > >
> > > I saw on the documentation:
> > > "That means all you CompSci graduate students who have been emailing
> > > me about doing your thesis on Nagios can contribute some code back.
> > >
"> > > I don't know who they are, but I try to do the job: change the
> > > algorithm of the circular check in order to have a O(n) complexity.
> > >
> > > My algo can optimise the circular check (for the moment, only the host
> > > part, but I'm working on the services part too). I use a personal
> > > modified version of the Deep First Search
> > > (http://en.wikipedia.org/wiki/Depth-first_search) (DFS)
> > >
> > > The hard work is already done: we've got link between parents->childs
> > > (the childs->parents link are not used). We "just" have to follow
> > > theses links.
> > >
> > > It's a recursive algo.
> > >
> > > All node are in the beginning unchecked (value =0), when we began to
> > > check them, it's temporary_checked (value=1). We check all childs if
> > > they are not already checked (here is the recursive part
).> > > If a child is in state temporary_checked, there is a loop, not follow
> > > the link, make our status loop_inside (value=3),and return it.
> > > If the child is a ok state (value=2), ok no problem with the child.
> > > If the child is LOOP_DETECTED (value=3), do not follow the link, and
> > > we return our status=loop_inside.
> > > If all childs are OK or if we don't have child (no childs, no loop
> > > ), we are OK an return.
> > >
> > > The algo is ok for code with #ifdef NSCORE (I need the host link).
> > > If we does not have NSCORE, we can't use it. But in the major system
> > > it's ok isn't it?
> > >
> >
> > Yes, the CGI's shouldn't ever read the un-expanded objects anyway, and
> > Nagios should never write circular parent chains to the objects.cache-
> > file, so whatever code there should be simply never kicks in.
>
> Ok
>
> >
> > > The modifications are in the files:
> > > *objects.h (to add macro DFS STATUS and add dfsCheckedStatus to host structure)
> > > *objects.c (to add DFS_UNCHECKED at the initialisation of host->
> > > dfsCheckedStatus)
> > > *config.c (add 2 fonctions: dfsStatus to get the status of a host, dfs
> > > for make the dfs algo for a host (and it's childs), and modifie the
> > > pre_flight_circular_check for call dfs on all node that need it (not
> > > already checked)).
> > >
> > > I know that the 2 function need to be in the objects.c, but I'll move
> > > them in the final patch.
> > >
> > > I generated a "big" conf to test it: 10000 hosts (milleservers.cfg),
> > > dependant of 100 parents (parents.cfg). The 10000 are not all looped.
> > > The loop are in the test.cfg file: 5 hosts, in a circular way.
> > >
> > > You can find the files (objects.c,h and config.c), the patch and the samples at:
> > > http://zegabes.free.fr/nagios/
> > >
> > > I put in this mail the code files and the patchs. The original files
> > > are from 2 hours ago.
> > >
> > > I try to make diff, but I don't know if they are ok with "patch"
> > >
> >
> > They are, b
...[email truncated]...
This post was automatically imported from historical nagios-devel mailing list archives
Original poster: [email protected]