[hobbit] purple recovery

[Henrik Stoerner]

On Mon, May 09, 2005 at 02:28:39PM -0400, Sue Bauer-Lee wrote:
> 
> I setup Hobbit Monitor only to find out that Hobbit coulnd't get a 
> netowrk response from a couple of systems. The networking issues have been
> resolved in that the server itself can get a ping/fping connectivity response
> from an interface that it's monitoring but Hobbit still reports that the 
> interface is unreachable and thus remains purple. 
> 
> 	System unreachable for 553 poll periods (173239 seconds)
> 			
> 		yet
> 
> /usr/sbin/fping 192.168.200.145
> 192.168.200.145 is alive

If it's purple, that means the status has not been updated for more than
30 minutes - so the ping-test is not being run for this host. Have you
changed the configuration to include a "noconn" for this host ?

> Speaking of connectivity, have I misunderstood the difference between
> "noping" and "noconn"? It seems that if I set "noconn", I still get pages yet
> "noping" produces the empty white circles and no more alerts. 

"noconn" completely stops reporting any status for the "conn" column.
So if you have this present from the beginning when a host is added to
the bb-hosts file, you will not have a "conn" column for this host.
However, if you start out with either a normal ping test or "noping"
(which just sends a "this test is disabled" status), then you have a
"conn" status column for this host - and then you must explicitly remove
that column if you add a "noconn" tag later on. If you don't remove it,
then it will go purple (because it is not being updated any more), and
purple statuses can result in alerts going out.

To remove the status column, run

    ~hobbit/server/bin/bb 127.0.0.1 "drop HOSTNAME conn"


Regards,
Henrik