Maximum runtime of 58 seconds exceeded for poller id 1. Exiting.

[tamanullah]

I am getting intermittent warning of 'Maximum runtime of 58 seconds exceeded for poller id 1. Exiting'.

I have resources mentioned below:
Method:spine Processes:4 Threads:20 Hosts:99 HostsPerProcess:25 DataSources:9877 RRDsProcessed:4988

On average Poller taker 4 to 5 minutes to process all this but sometime it takes more than 58 seconds.
Dry run of spine is not resulting any useful result as issue is not consistent.
Another strange thing is while poller reports runtime of more than 58 seconds but in devices section, I do not see any device taking abnormal time.

I have updated cacti from 1.2.21 to 1.2.23 but still no result. I also enabled boost but the frequency of warnings did not come down.
This is resulting in missing graphs on exact time when warning is issued.

Can anyone help me identify the issue or can point out what I can check?

[macan]

Data collection->data collectors - try to encrease threads for poller.
Try to increase/decrease ping and snmp timeouts and retry count.

Sort your devices by poll time and try to resolve the worst devices