Improving dependencies mechanics ?

[Folkerd Koning]

My advice on #2 would be to force a check on depending services immediately
after a state change for dependant services (possibly as a configuration
option in the main config file or on a per-service or per-host basis). That
should update it's services correctly. Kicking them to "unknown"
instantaneously could create notifications that aren't neccesary.

First checking dependant services prior to checking the depending service
(or host) could create a LOT of checking for certain hosts or services. Say,
you have 15 SQL servers on a single NAS, and you check those servers for
load every minute. That would cause the NAS to get checked 15 times per
minute! These kinds of checks I would not recommend, especially when there
is no state change in the depending service. If the depending service
changes state, then I would also kick off a forced check on the dependant
service (IE on state change on either dependant or depending service, check
the whole chain).


Regards,
  Folkerd H. Koning
  Email Administrator 
  City of Fort Wayne,Indiana
  County of Allen, Indiana
  Tel: (260) 427-5166


 

> -----Original Message-----
> From: nagios-devel-bounces at lists.sourceforge.net 
> [mailto:nagios-devel-bounces at lists.sourceforge.net] On Behalf 
> Of Laurent Guyon
> Sent: Monday, September 04, 2006 10:48 AM
> To: nagios-devel at lists.sourceforge.net
> Subject: [Nagios-devel] Improving dependencies mechanics ?
> 
> Hello,
> 
> I've got a question about dependencies system in Nagios.
> 
> If I understood well the actual mechanics, if service A 
> depends en service B, and if service B goes into trouble, 
> checks on service A are stopped. But I noticed 2 things :
> * Services test order is not affected by dependencies informations,
> * Service A remains in its current state when B goes into trouble.
> 
> I would like to improve this system, to add a notion of 
> precedence, and a system of dependent services state correction :
> 1) As service A depends on service B, nagios must first check 
> service B, and then check service A if service B was OK
> 2) If service B fall into trouble (CRITICAL or WARNING), 
> correct service A state to UNKNOWN.
> 
> I understand that the notion of precedence would be hard to 
> set, because of the differences of services poll intervals 
> (if service B is checked every 2 minutes, and service A every 
> 1 minute ...).
> 
> I've tried to achieve 2) with event handlers, when service B 
> goes into trouble, an event handler corrects all its 
> dependants services (A) to set them to UNKNOWN. But it's not 
> perfect in a certain case : if a normal check of service A 
> have been launched just before service B goes into trouble, 
> the result of this check returns AFTER service A state have 
> been corrected to UNKNOWN, and overrdide its UNKNOWN state.
> 
> Do you know how I could do that, or have you any other idea 
> to improve the dependencies system ?
> 
> Thanks in advance,
> 
> Laurent
> 
> 
> --------------------------------------------------------------
> -----------
> Using Tomcat but need to do more? Need to support web 
> services, security?
> Get stuff done quickly with pre-integrated technology to make 
> your job easier Download IBM WebSphere Application Server 
> v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&
> dat=121642
> _______________________________________________
> Nagios-devel mailing list
> Nagios-devel at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/nagios-devel


-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642