[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [hobbit] Some configuration questions



On Tue, May 30, 2006 at 11:16:59AM -0500, Hubbard, Greg L wrote:
> Hobbits:
> 
> Some general questions, based on 4.1...
> 
> A) Do you know if the keywords in bb-hosts are case sensitive?  I
> noticed that the documentation is very careful to match case in examples
> as well as the definitions -- but some keywords are in ALL CAPS and
> others are lower case.  One wonders why this is so...unless they are
> case-insensitive, and it is just a curiosity in Hobbit-lore...

It's mostly historical, based on the Big Brother ancestry.

I don't have the full picture in my head, but in most cases it is not
case-sensitive. The "bbgen" and "bbtest-net" are slightly more picky,
for historical reasons - they still do some parsing of the bb-hosts
file by themselves. Specifically, I know the "depends" and "route"
tags are case-sensitive.


> B) Regarding the "route" keyword?  Can these be chained, or does the
> whole path have to be specified for each item in the file?
> 
> Example:  hobbit server -- firewall -- router a -- router b -- host --
> app
> 
> Can it be coded this way?
> 
> Firewall 	no route statement
> Router A	route:firewall
> Router B	route:router A
> Host		route:router B
> 
> Or do I have to say this:
> 
> Router B	route:firewall,router A
> Host		route:firewall,router A,router B

The latter one will definitely work. I am not sure if the first one
will; a quick glance at the code appears not to support these tree-like
dependencies.

It would be nice to feed a map of your network topology into Hobbit
so when some router breaks, it will automatically know what tests 
will go red as a result. It's just a little difficult to implement, 
I think. Especially how you would describe your network topology.


> C) I have some custom tests that are run by the client on "Host" --
> these messages (which go purple) don't seem to get suppressed if a node
> is unreachable because of something else.  Is there some way to code
> this to take advantage of the downstream suppression?

If you have a "conn" test for that host, they shouldn't go purple.
<checking the code> wait, that only works if the "conn" test is
red,yellow or blue - not clear, which is what the dependency checks
make it. Will fix.


> BTW -- Henrik -- I have seen the downstream suppression stuff work --
> awesome!  It is a lot more difficult to achieve this in my commercial
> tools.


Thanks,
Henrik