templating

[bmoyles]

Okay, I'm probably missing something here (I'm sure data scripts could help in some way, but I've got coders block today )

We've got a Cisco/Arrowpoint CSS load balancer. We want to track traffic for each virtual IP (corresponding to a particular load balanced site).

Information in the CSS is grouped by Owner, Content, and then Server. Owner is sort of like a user. Content is the actual rule that defines the virtual IP. Servers are virtual servers you add to content rules--they make up the pool.

Okay, that said, the CSS, instead of using indexes to store the counters for these virtual interfaces actually encodes the Owner, Content, and Server names in the OID itself. Funky.

For example, lets say my owner is Ignite, and my content is called icecast.

Here's the OID you'd use to pull the bandwidth counter for that:
1.3.6.1.4.1.2467.1.16.4.1.25.6.73.103.110.105.116.101.7.105
.99.101.99.97.115.116

.1.3.6.1.4.1.2467.1.16.4.1.25 is the base
.6. is the length of the next encoded string
.73.103.110.105.116.101. is Ignite encoded in ASCII (6 characters long)
.7. is the length of the next encoded string
.105
.99.101.99.97.115.116. is icecast encoded in ASCII

Fun.


Okay, that said, I can make a host template for the CSS. I can make a data template that lays out how I want the data to be collected. I can set the OID to Use Per-Data Source Value (Ignore this Value), but that means I have to create a new datasource for each content rule, and means I can't use graph templates.

Any insight on how I might go about automating this a bit more?
Thanks in advance,
Brian Moyles
Sportvision, Inc.

[raX]

From what I gather, you are looking for a way to automatically query a list of servers attached to your CSS load balancer. This is exactly what data queries are for except your configuration is a bit different...

Data queries currently require each row to have a single unique index value. Currently this value cannot span more than one octet in the OID based on how SNMP works.

On the other hand, this might be possible using a script query. You could theoretically write a perl/PHP script to do the SNMP polling for you and generate a unique index for Cacti's use only. If you are interested in trying this, I recommend that you check out 'scripts/query_host_partitions.php'.

About potential data/graph template configuration, it really depends what data the CSS can give you. Is the monitorable data specific to the 'content type' in use? If so, I would recommend data/graph templates for each content type. On the other hand you could probably use a single graph/data template to serve all servers on a given CSS.

-Ian

[bmoyles]

Hmm, I'll look into the script queries. I briefly glanced at them when I upgraded to .8, but didn't give them much of a look. At second glance, it seems like they're almost exactly what I need to do. Thanks!