"Blocky" graphs ? Update intervals fixed to 60s ?

[picpic]

Hello,

I have set up a Zabbix server and everything is working great ! I didn't have any problems at all. Very intuitive interface, basic templates fulfill all my needs.

I'm currently monitoring my server with SNMP. I need to monitor the CPU load for some of them. So I manually created CPU load graphs for those servers.

Since I'd like my CPU graph to have a 15s update interval between values. Here's what I did:

- In "Template SNMP Processor" --> "Item Prototypes" --> "Utilization of processor #{#SNMPINDEX}", I changed the "Update Interval value" from 60 to 15 seconds. The item was correctly updated on all the servers using this template.
- Then in Host configuration --> [my host] --> "Graphs", I created a graph called "CPU" and manually added the different cores.

The graph is correctly generated for the desired host. The only problem is that, despite the value update time set to 15s, it looks like I only have values updates every 60 seconds on the graph. Which gives this "blocky" look. See below:



Is there a way to have shorter updates intervals ? Is modifying the item update interval the way to go ? Or am I missing something ?


Thanks ;-)
Have a nice day !

[picpic]

Ok...

After resetting the update interval of the processor item to the default 60s instead of 15s in the SNMP Template, my graphs looks normal again !



Still, the update interval is 60s. And i'd like 15s.

My question is: does the "Update Interval" parameter actually controls the SNMP polling interval ?

It looks like it only reads the SNMP data, which is only updated every minute. Hence the blocky graphs if I set the interval to 15s (it will read 4x the same value in 1 minute).

Is SNMP polling actually limited to 1 minute intervals ?

[jan.garaj]

It looks that your SNMP value is updated once per 30 seconds - that's the reason why 15s graph has blocks (2 same values per 30sec), but 60s graph doesn't have blocks. You can poll SNMP load value even every second, but value will be updated/calculated only once per 30sec. Your best polling interval, which make sense is 30 sec. Check your SNMP cpu implementation/manual :-)

Or you can use zabbix agent and then you can poll (or push) CPU load even every second without any problem. I use 1 second interval for special load tests - active mode is recommended - see http://i.imgur.com/GZ5mSF8.png http://i.imgur.com/TeILnML.png.

Zabbix agent implements "collector" for CPU statistics, so it doesn't depend on (old/broken) SNMP implementation.
Keep in mind that Zabbix CPU load1/5/15 (+usage times) is avg value calculated every second for last 1/5/15minutes.

If you need to know also values for different CPU times (system/user/iowait/nice/...), then your choice must be Zabbix agent.

[Alcheemy]

Unless I am mistaken and are thinking about something else, I believe I got the solution for you.

I was looking for this myself awhile back, and struggled to figure out how to update the graph interval. I am assuming you want the graph to update itself every x seconds, which is a different update timer than the one in the "Item settings".

If you go to your "Profile" in upper right corner, and access your profile-settings you can edit the field "Refresh (in seconds)". Put the value to whatever x you want for the graph to update at.

[picpic]

@alcheemy: thanks, but i'm talking about the values in the graph & how they are displayed, and not about the refresh of the page itself which does not have any effect on the data displayed

@jangaraj: You are saying that snmp can retrieve values down to every second. Well, that's exactly what I need ! But I can't find the way to do it. Could you please point me to this "snmp/cpu implementation manual" you are talking about ? I can't seem to find this information anywhere.

Thanks !

[picpic]

For those interested: you can have more time-accurate cpu graphs using the active agent and querying windows processor perf counters. I used the total processor perf counter as an item with 10s Refresh intervals, so now I have pretty accurate total cpu graphs in %.

You can search for perf counters in the zabbix doc, it's pretty straightforward. Keep an eye on your server load & queues, though. Maybe you'll need to adjust some parameters client-side & server-side. But that's another story.

Have fun !