Cisco T1 Controller Counters

[siouxee]

I also noticed that the poller cache shows the same index OID for all of the different queries instead of the actual OID for each counter. Will a bogus OID (an index OID in this case) return a value of "-1"? How do I get the correct OIDs queried?

[BSOD2600]

Post the cisco_ds1.xml file.

Also follow http://docs.cacti.net/manual:087:4_help ... #debugging to troubleshoot.

Your Data Sources (i.e. cacti_data_source_debug_051509.JPG) has the correct index number, so you're on the right track.

Did you forget to check the boxes in the Data Query under the Associated Data Templates section?

[siouxee]

Current cisco_ds1.xml file is attached. The check boxes are checked for data queries and associated graph templates.

Did you see my previous post about the poller cache showing the same index OID for everything instead of the individual OID for each counter? Any idea why that would be happening?

[BSOD2600]

1) you know your screenshots have the snmp community?
2) Yea, saw that screenshot. The cisco_ds1.xml appears correct. Clear the poller cache -- see if that fixes those OIDs.
3) I see each oid type has a separate data rrd file, which means a unique data template for each? Any reason why you don't have them all in the same logical data template?

[siouxee]

1) thanks for pointing that out, but got bigger things to worry about than someone getting my read-only string, which is only accessible internally... it's still version 1 so it's not like it's secure to begin with.

2) already tried clearing the poller cache, no change.

3) would this make a difference with it polling the wrong OIDs (or in this case polling only the index OIDs), other than keeping them together?

[siouxee]

OK, I just noticed that in my current screen capture of the poller cache (after clearing it a couple of days ago) the individual counter OIDs that are being queried are the correct OIDs, but the last four fields (.9.1.1.11, .9.1.1.12) of the index OID are getting appended to the end of them instead of just the last field (.11, .12) which represents the controller being queried. Hence, these are not returning valid values and the errors are being generated. How are these extra fields being appended to the query?

Could it have something to do with the following line in the XML file?

<oid_index_parse>OID/REGEXP:.*\.([0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3})$</oid_index_parse>


Should it say something like this instead to only append the last field value of the index OID?

<oid_index_parse>OID/REGEXP:.*\.([0-9]{1,3})$</oid_index_parse>

or this: <oid_index_parse>OID/REGEXP:.*\.[0-9]{1,3}$</oid_index_parse> (not sure if the parentheses are needed or not)

For my example, the OIDs being queried should look something like this:

.1.3.6.1.2.1.10.18.9.1.9.11
.1.3.6.1.2.1.10.18.9.1.2.11
.1.3.6.1.2.1.10.18.9.1.11.11

Instead of this (currently in the poller cache):

.1.3.6.1.2.1.10.18.9.1.9.9.1.1.11
.1.3.6.1.2.1.10.18.9.1.2.9.1.1.11
.1.3.6.1.2.1.10.18.9.1.11.9.1.1.11

[siouxee]

I've gone through the docs again and have simplified the XML file. I'll test and see what happens but the poller cache is showing the correct OIDs now! I'll post everything once I confirm everything is good.

Back to the data template, does it make more sense to combine everything under one template to keep things orderly? Does this speed up the polling process, etc?

[BSOD2600]

Back to the data template, does it make more sense to combine everything under one template to keep things orderly? Does this speed up the polling process, etc?

Personally, I like to keep things logically grouped together the same data source (DS) items in my data template as the associated graph template. Yes, there is a tradeoff point too -- one DS per rrd file vs 203345 DS's per rrd file. I'm sure there is some performance impact on that, but that question is better suited for the rrdtool mailing list.

Being that you've only got a dozen DS's, I'd put them all in the same data template.

[siouxee]

It appears the data is getting queried properly (see latest attached XML file) but now the graphs are displaying some crazy numbers. Is there an easy way to see what value is being returned when the query executes?

I currently have the data template "Data Source Type" set to "COUNTER" since this value will always increment, unless I manually go clear the counters on the router. Not sure if I'm in the right ballpark on the cause of the graph having the wrong values or not...

Here's the current SNMPGET for this graph in question, with current value of "2505":

snmpget -c R!t3$HsUk$! -v 1 10.4.10.100 1.3.6.1.2.1.10.18.9.1.2.11
SNMPv2-SMI::transmission.18.9.1.2.11 = Gauge32: 2505

And, here's the Cacti debug output for a manual polling interval, showing the correct value of "2505":

05/17/2009 07:34:28 PM - CMDPHP: Poller[0] Host[7] DS[2505] SNMP: v1: lkc-ld-08-ar01.englobal.com, dsname: dsx1TotalESs, oid: .1.3.6.1.2.1.10.18.9.1.2.11, output: 2505
05/17/2009 07:34:28 PM - CMDPHP: Poller[0] DEBUG: SQL Exec: "insert into poller_output (local_data_id, rrd_name, time, output) values (2505, 'dsx1TotalESs', '2009-05-17 19:34:28', '2505')

[BSOD2600]

It appears the data is getting queried properly (see latest attached XML file) but now the graphs are displaying some crazy numbers. Is there an easy way to see what value is being returned when the query executes?

Change the cacti logging level to medium or better.

You'll want to read over http://forums.cacti.net/viewtopic.php?t=8168.

You're likely seeing those weird numbers due using a counter and rrdtool averaging the data over a 5 min period.

[siouxee]

Here's the Cacti debug output for a manual polling interval, showing the correct value of "2505":

05/17/2009 07:34:28 PM - CMDPHP: Poller[0] Host[7] DS[2505] SNMP: v1: lkc-ld-08-ar01.englobal.com, dsname: dsx1TotalESs, oid: .1.3.6.1.2.1.10.18.9.1.2.11, output: 2505
05/17/2009 07:34:28 PM - CMDPHP: Poller[0] DEBUG: SQL Exec: "insert into poller_output (local_data_id, rrd_name, time, output) values (2505, 'dsx1TotalESs', '2009-05-17 19:34:28', '2505')

I'll check out the link you provided...

[BSOD2600]

Being that all those OIDs you're tracking in the DS1-MIB file are type GAUGE -- you should be using that in the data template too. A counter means the value increments / changes with time (like traffic data). Once you change them (and set reasonable MAX DS limits!), you'll need to delete the rrd file(s) so Cacti re-creates them with the new DS types.

[siouxee]

Yeah, figured out the GAUGE part last night and the graphs are working properly now! I'm working on cleaning up based on your recommendations and will probably consolidate all of the different error counters into one graph per T1 controller (since these graphs should never have much constant data on them, assuming the circuits are behaving normally). Still debating that one, since some of these error counters could stay in the hundreds, while some could jump into the millions at any given time (graph data may be hard to see for lower values). Any suggestions or should I just keep each counter on it's own graph? We're talking ten or so seperate graphs per T1 controller if I keep them seperate.

Do you recommend setting a "0" / unlimited max value for the DS? That's what I've been doing for this one. The counters could go into the millions, or could stay in the thousands, just depends on the severity of the circuit's problems.

[BSOD2600]

Any suggestions or should I just keep each counter on it's own graph? We're talking ten or so seperate graphs per T1 controller if I keep them seperate.

One has to look at the 'average' data and decide if various counters will fit logically on the same graph or if one will sqew the picture. I really is up to the end user how they want to monitor, display and alert on the data. I'm not privy on the the data you're collecting, but possibly consider creating a 'stacked' graph -- only if the counters can be added to each other.

Do you recommend setting a "0" / unlimited max value for the DS?

0 is generally ok for GAUGE data sources. Might want to look in the DS1-MIB file and see if there is a max defined for the counters though. For a COUNTER data source, one wants to always set an appropiate MAX.

[jgarland79]

I just finished making my own custom templates for monitoring DS1 error counters. See attached. It could use a little more tweaking if anyone is up for it.