Poller overruns after 1.2.23 upgrade

[KevinNoall]

I've done the 1.2.23 upgrade on a couple different systems, but on my main install my poller runtime went from ~12-20sec to 58+ and gets killed regularly.
Centos 7x, PHP 7.4.x, 10.5.15-MariaDB
I tried disabling the few plugins I had running to no avail (current thold, monitor, webseer)
I've seen some errors about mysql db going away and have seen mysql fail. Not exactly sure what would have changed to cause this trouble?

[TheWitness]

Grab the latest spine. There were some issues with the 1.2.23 release, fixed now. Also, if upgrading from 1.1.x or 0.8.x to 1.2.23, the threading and processes moved to Data Collector > Edit.

[KevinNoall]

This seems to have resolved after grabbing the latest spine. Thanks

[KevinNoall]

After upgrading to the new cacti 1.2.24 this issue returned. I tried putting spine back to v1.2.22 but that did not resolve. all remote pollers and the main poller are all overrunning the polling time and collecting no statistics.
Nothing alarming in the logs that I can see

[KevinNoall]

Now I'm seeing boost errors in the log:
BOOST ERROR: Failed to retrieve any rows from archive tables
and getting emails:
WARNING: Detected Poller Boost Overrun, Possible Boost Poller Crash

[TheWitness]

Upgrade to 1.2.24 ASAP.

[KevinNoall]

Apologies for re-using the post. this is AFTER I upgraded everything to 1.2.24

[TheWitness]

Is it happening often? Make sure there is at least an hour between boost updates.

[KevinNoall]

I'm getting the warning emails every minute.

The setting was 30m, I've changed it to 1h.

all pollers dropped back to spine 1.2.22 with no improvement.

[TheWitness]

How big is this install?

[KevinNoall]

15 remote pollers, but not all that many devices overall about 300

[KevinNoall]

Even with remote pollers all disabled, the main poller time out and it only has ~20 devices

[TheWitness]

How about the server? CPU cores/threads? How much memory?

How many collector processes and threads in total?

[KevinNoall]

2gb RAM, 3 processors. remote pollers the same

3 processes 20 threads per process for each poller

avg polling times were in the neighborhood of ~30seconds for all the pollers prior to 1.2.24 upgrade.

[TheWitness]

With what you're saying you really need considerably more threads on the core or the main server are they VMS if they are go to 16 cores and well more memory for sure go to like 8 gig memory at a minimum.