Balance Process Load does not always apply properly

[chpoqxie]

I have high devices' CPU snmp load after upgrade to 1.2.23.
When I tried to figure it out, I found out the following:

Poller settings:

Poller log before upgrade (rrds are balanced between cron works):

Poller log after upgrade (every cron work ALL rrds are processed):

any other info by your request if needed

[chpoqxie]

also I can't set "Poller Interval" to 5 minutes as configured in used "Data Profile".

[TheWitness]

Three recommendations:

1. Switch to systemd cactid service and move away from cron
2. Switch cron/service interval to 5 minutes, and poller to 1 minute
3. Repopulate the poller cache.

[chpoqxie]

1. i dont have systemd because use freebsd ) does the cactid work under freebsd?
2. why switch cron to a 5-minute interval if I need a balanced every minute work? I don't understand this recommendation, it does the opposite of what I require.
3. done.

i think you recommends not that i want to get from cacti.
after steps 2 and 3, I get the same number of rrd-files processed every minute and an every minute SNMP poll of the same devices with the same values (checked by tcpdump):

i have used every 5 minute data source profile, not every 1 (that's why I want poller to work as before - so that the number of requests and rrd-files is balanced between cron tasks):

[TheWitness]

You mis-understand 'Balance Process Load' is to distribute the hosts by the number of Data Sources, not to spread out data collection. You should set your Cron interval to 5 minutes and your Poller Interval to 1 minute and then re-populate the poller cache.

[chpoqxie]

to prevent misunderstanding, please specify where I need to change the "Poller Interval"?
at the Configuration - Settings - Poller, or at the Presets - Data Profiles? in the case of the first option - I did it before the previous comment.

previously, the Cacti worked in such a way that if i selected the "5 minutes" Data Source Profile and set the Polling and Cron Intervals in Settings "every minute" (and set "Balance Process Load" on, perhaps it was not necessary), all the requested data was evenly spread over the poller launches through cron every minute.

now after the update it stopped happening, ALL data requested and ALL rrd files writed to disk every 1 minute, not every 5. i dont need so many data )) and some devices are overloaded in this case.
how to return the old state in the current version of the program?

[TheWitness]

Have you followed my instructions yet?

[TheWitness]

Okay, confirmed this now after testing your setup.

[chpoqxie]

thank you very much )
i applied this two patches to my cacti installation ("Polling distribution not working" and "Polling Issues"), touched "Rebuild Poller Cache" again, and see this progress in logs:

it now works the way i wanted )

[TheWitness]

I did make one more change. Please take note. It was done this evening.