Gaps in Graphs

[newmanity]

I have recently added an extra 8 graphs to 45 devices, each of which has a PHP script to go through (it pulls an OID from one device once it has found it's MAC in the OID list on another). I have a fairly slow machine doing this work: it has a Pentium 4 2.8 GHz processor with 1GB RAM.

I am now noticing that some of the devices are displaying gaps in their graphs, and as the all the devices are the same and are using the same templates, I figure this is a problem with too many things being run by the poller. I have set the php allocated memory to 128M and the timeout is 30 seconds. I am getting the following types of errors in the log file:

Code: Select all

CMDPHP: Poller[0] DEBUG: SQL Exec: "insert into poller_output (local_data_id, rrd_name, time, output) values (673, 'apRssiDSI', '2008-04-15 14:15:17', 'U')"
CMDPHP: Poller[0] Host[36] DS[673] SERVER: /var/www/cacti/scripts/ss_Canopy.php ss_Canopy 10.5.12.127 localnet rssi, output: U
CMDPHP: Poller[0] Host[36] DS[673] WARNING: Result from SERVER not valid. Partial Result:

I have tried playing with the max concurrent poller process, setting values from 1 to 20, with no success. I read that spine is a lot quicker, and so I tried switching to spine using the recommended range values (I have set the spine parameters, in order, to: 10,7,30,10 at the moment). Spine doesn't appear to be any better. I get the following errors in the log file:

Code: Select all

SPINE: Poller[0] Host[49] DS[560] SS[1] WARNING: Result from SERVER not valid. Partial Result: ...
SPINE: Poller[0] ERROR: SS[1] PHP Script Server communications lost.

Any clarifications or recommendations would be most welcome.
Thanks.

[gandalf]

Spine settings are not as recommended. Use 1-2 processes, 10-15 threads and 2 script_servers as a start.
What is the runtime of that script you've mentioned?
Are you aware of the fact, that there are already a lot of Canopy resources on the forums? Did you verify, that your problem was not already solved by someone?
Reinhard

[newmanity]

That doesn't seem to improve things, and yes, I've tried looking in the forums for Canopy related posts for this issue. The thing is, I'm using my own scripts at the moment, and I simply want to figure out whether or not the gaps I'm getting are a result of the network or the processing power of my machine.

I am trying to display data relating to a given wireless canopy receiver (SM) on a single page. Some of the stats that I'd like to show, however, are stored on the transmitter side (AP). To access the OID, I have to identify a dynamically assigned link ID (LUID) to that SM which is added as the last number of the OID string. This means that I have to SNMP the SM to get the SMs MAC and its associated AP's MAC (to get it's IP). I then have to snmp search through the AP's SMs to get the LUID. Once I have the LUID, I can snmp to get the value I'm looking for. I'm not sure if this will create too much traffic on the APs or not. I have also tried pulling the LUID from a static list from a table to reduce the load, but this has had no impact whatsoever.

Is there any way to identify whether this is my own machine that is slowing things down, or whether the APs themselves cannot handle all these snmp requests? Would I get a specific error over and above the partial result error?

Thanks.

[gandalf]

As those devices are not that fast, I fear that the runtime of your script may exceed script runtime limits. Personally, I did not bother with the SNMP traffic. But you may use the SNMP traffic template that's part of the Template provided by BSOD2600 named "MIB Protocol Stats"
Reinhard