Sangeeta Misra wrote:
>Folks,
>
>We are in need of the following server monitoring capabilities in
>Solaris loadbalancer:
>
>o ping probe where a ping is sent by the loadbalancer
> to each backend server
>o TCP probe- where a TCP three-way handshake with a server is
> performed on a specific port
>o UDP probe to see if the UDP stack on the server is listening
> on the speccific port
>
>
>I am wondering if anyone has used OpenNMS in Solaris, and is aware of
>whether it can be used for the aforementioned probes. It looks like
>OpenNMS includes ping probe, but I am not sure if the TCP/UDP probes
>have already been implemented or not.
>
>Also are any there opensource software that folks recommend for the
>above use?
>
When I played with this back in 2000, I'm not sure if there was
an OpenNMS, but it really wasn't that hard to do...(famous last
words, I know...)
There is one other probe that you haven't listed that I would
consider important and that is being able to send a query to
a TCP port and match it against a reply. For example, being
able to send "GET /systemhealthcheck HTTP/1.0" to the web
server and then comparing what's received, if it is "response
code 200", with the same file locally.
It may also be worth adding the concept of a "link probe",
where a loadbalancer is directly connected to what it is
trying to manage.
So you might have:
link probe - am I still connected to what it is I'm redirecting to
ping probe - is the address/kernel of the remote machine working
tcp probe - is the application port open
content probe - is the application able to respond to static requests
php probe - is the server able to answer "complex" queries that may
involve a database
I know you haven't listed the latter two and I think that's a bit
of a bug. You should at least consider having a "content probe".
I think that using OpenNMS in this way is the wrong way to think
about this as it brings with it a lot of baggage. A better question,
with respect to OpenNMS, would be to consider how would OpenNMS be
aware of such a daemon and/or know that the daemon is active/faulty?
Darren
_______________________________________________
networking-discuss mailing list
[email protected]
