WARNING: Cacti Polling Cycle Exceeded Poller Interval by 2.35 seconds

[mohammadramadan]

Hello,
I am receiving this "WARNING: Cacti Polling Cycle Exceeded Poller Interval by 2.35 seconds" after setting the "Poller Interval" to be every one minute.
Do I need to take an action? what is it? I set it to have a faster thold response.
check the attached "thold plugin" setting.
my vitual maching is ubuntu linux 22.04, 4 cpu single core each, 8 gig RAM., my devices added 300 snmp devices.

[cigamit]

Show the SYSTEM STATS line from your cacti log. Most likely you just need to increase the processes / threads.

[cigamit]

BTW, you don't increase the processes there in Settings (thats the old place, and now is just the default for new pollers), you do it under Data Collection on the Console menu.

[mohammadramadan]

Hi,
my System State sample lines:
2023-08-24 22:16:03 - SYSTEM STATS: Time:2.0541 Method:cmd.php Processes:10 Threads:1 Hosts:91 HostsPerProcess:10 DataSources:0 RRDsProcessed:0
2023-08-24 22:15:03 - SYSTEM STATS: Time:2.0468 Method:cmd.php Processes:10 Threads:1 Hosts:91 HostsPerProcess:10 DataSources:0 RRDsProcessed:0
2023-08-24 22:14:08 - SYSTEM STATS: Time:2.0398 Method:cmd.php Processes:10 Threads:1 Hosts:91 HostsPerProcess:10 DataSources:983 RRDsProcessed:0
2023-08-24 22:12:03 - SYSTEM STATS: Time:2.0297 Method:cmd.php Processes:10 Threads:1 Hosts:91 HostsPerProcess:10 DataSources:983 RRDsProcessed:0
-----------------
I am trying to speed up the thold notification process because it takes from 7 to 9 minutes to get notified.
Check the attached image for the poller data setting.

[Osiris]

Switch to spine.

[cigamit]

Are you talking about the up/down alerts (could have sworn that was moved to Monitor) or the breaching of data points alerts?

Thold (breaching data points) doesn't get faster if you speed up the poller, as your graphs are still polling at the old rate, you are just spreading them out over the 5 minute period. Its your Data Profiles that decide how often the data in a graph is updated, and thold only checks each graph when it is updated.

So really, what you are wanting to do is move your Profiles from 5 minute data collecting to 1 minute data collecting. That is a bit more complicated, but there are various forum topics on here showing you how to do it.

[mohammadramadan]

Switch to spine.

In the settings I have CMD.php option only, my previous cacti used the spine and was fine but I didn't find it in this newer one, if you think switching to spine will solve it, so I will google it.

[mohammadramadan]

Are you talking about the up/down alerts (could have sworn that was moved to Monitor) or the breaching of data points alerts?

Thold (breaching data points) doesn't get faster if you speed up the poller, as your graphs are still polling at the old rate, you are just spreading them out over the 5 minute period. Its your Data Profiles that decide how often the data in a graph is updated, and thold only checks each graph when it is updated.

So really, what you are wanting to do is move your Profiles from 5 minute data collecting to 1 minute data collecting. That is a bit more complicated, but there are various forum topics on here showing you how to do it.

Ok, then, I reverted it back to the 5 mins, and now searching for converting profile from 5 min to 1 min.

[Osiris]

Spine is a c binary the is Uber scalable. Switching from 5 minute to one minute is best handled by a trained cacti professional, but these days it's much easier to just reinstall.

If you don't care about the old data, you can do 7 things to be switched.

1) Switch all the Data Templates to Use the 1 Minute profile
2) Switch the Poller to one minute poller
3) Run one database alter command to switch the data sources to one minute profile
4) Switch to spine
5) Switch to Cactid from cron
6) Delete all the rrdfiles
7) Repopulate the poller cache

[Osiris]

It's always recommended before you go too deep into an install to plan things like

1) How much history do I want, and at what aggregation levels
2) What data collection intervals will you support for each graph type.

[mohammadramadan]

Spine is a c binary the is Uber scalable. Switching from 5 minute to one minute is best handled by a trained cacti professional, but these days it's much easier to just reinstall.

If you don't care about the old data, you can do 7 things to be switched.

1) Switch all the Data Templates to Use the 1 Minute profile
2) Switch the Poller to one minute poller
3) Run one database alter command to switch the data sources to one minute profile
4) Switch to spine
5) Switch to Cactid from cron
6) Delete all the rrdfiles
7) Repopulate the poller cache

Hello Bro,
sorry for the late responce as I was in cretical datacenter status, and here I am back to my beloved cacti
Well..
1- I have now about 130 Device "considering all of them with snmp enabled" what is the cacti appliance recommended requirements "CPU,RAM, and HDD" if this would help avoiding this subjected issue?
2- the issue occures just randomely, I mean while every thing is ok, it just get crazy and start sending this notification, sometimes after network major issue, and sometimes else after no network change.
Restarting the server solve the issue. so what is the service the cold be restarted to solve this? and how to avoid this issue in general?
3- If you consider spine is better for my case, I wouldn't mind to reinstall the whole server, but I am still unable to find the latest cacti with spine version and have it compatible with at least the thold, Monitor, Mactrack, and Intropage.

kindly reply answering my questions.
TIA

[macan]

use version 1.2.25 from https://www.cacti.net/

[mohammadramadan]

Hello,
Thank you so much, I was surprised that I had issues with version 1.2.24 but not it was few days ago that spine ver 1.2.25 available.
I have installed it and really appreciate the whole team work.
Almost all my previous problem solved. still testing it.
Thank you again,